anyone else getting this error: svn: Berkeley DB error while opening 'nodes' table for filesystem /var/svn/gnue/db: Cannot allocate memory bleh hey chillywilly are you the author of dcl? nope michael dean is mdean on IRC he's not around much ah neat *** reinhard has joined #gnuenterprise *** johannesV has joined #gnuenterprise *** dcmwai has quit IRC *** Amorphous has quit IRC *** derek has quit IRC *** holycow has quit IRC *** nickr has quit IRC *** holycow has joined #gnuenterprise *** nickr has joined #gnuenterprise *** reinhard has quit IRC *** chillywilly has quit IRC *** dcmwai has joined #gnuenterprise *** Amorphous has joined #gnuenterprise *** derek has joined #gnuenterprise *** reinhard has joined #gnuenterprise *** chillywilly has joined #gnuenterprise *** btami has joined #gnuenterprise *** SachaS has joined #gnuenterprise morning all http://wizards-of-os.org/index.php?id=835 thats in Berlin in a couple of days hi SachaS hi btami mornin SachaS hi btami cool linkage SachaS, where is your baby erp project writeup? the planning stuff? i lost the bookmark from earlier, would love to read that again holycow babyerp is broken you have to get it from svn and there is a readme in the source ahhhh okay my svn still throws that error screenshots when it was still working: http://www.gnuenterprise.org/~sachas/screens/ wow *** holycow has left #gnuenterprise *** holycow has joined #gnuenterprise that is awesome perhaps i'm easily impressed... but i can see that alone doing some serious work we gotta fix that puppy up! there are no real concepts behind *nod* i realize this is an excercise in gnue, but it looks quite workable for a small business sorry, just so that you know, i run lotus notes and domino at our company right now... which is designed as a forms based environment, amongst other things, this looks v v good in comparison * holycow kicks the svn server damnit, i gotta get gnue soon :) ehe times a wastin hi holycow the latest you can get here: http://www.gnuenterprise.org/downloads/snapshots.php the svn access is down again, sigh i understand if i get them from cvs... they will work in the local download directory? ahhhhh k. was curious k. gotta get some sleep. nite. thx for the linkage np *** dimas has joined #gnuenterprise the newest debian installer is so slick *** kilo has joined #gnuenterprise *** johannesV has left #gnuenterprise *** SachaS has quit IRC *** SachaS has joined #gnuenterprise * SachaS went to enemy territory gnucash. they have a link to gnue so thats cool goooood morning hi kilo you already awake??? again. brrrrrr *** holycow has quit IRC *** SachaS has quit IRC *** SachaS has joined #gnuenterprise *** kilo has quit IRC *** SachaS has quit IRC *** kilo has joined #gnuenterprise *** SachaS has joined #gnuenterprise *** SachaS_ has joined #gnuenterprise *** SachaS has quit IRC *** btami has quit IRC *** johannesV has joined #gnuenterprise *** dsmith has joined #gnuenterprise *** Vee2d2 has joined #gnuenterprise *** mixi^ has quit IRC *** kilo has quit IRC *** reinhard_ has joined #gnuenterprise *** reinhard has quit IRC *** jamest has joined #gnuenterprise *** kderichs has joined #gnuenterprise *** kderichs has quit IRC *** btami has joined #gnuenterprise jamest: svn: Berkeley DB error while opening 'nodes' table for filesystem /var/svn/gnue/db: Cannot allocate memory t *** mixi has joined #gnuenterprise geez * chillywilly kicks the gnue svn btami: i think we could install a bot that automatically posts these lines as soon as jamest logs in There is something screwey going on. ;-) heh hmm, and the latest snapshot is 20040602 anyone try the latest incantation of gthe debian installer it is sweet reinhard :) chillywilly: Yeah? yep detects a lot of things for you if latest = last beta, then yes even has automagic partition (which you can edit later) partitioning chillywilly: yes, it's great it's really great reinhard: release candidate 1 which ~= beta 4 ok i think they counted it indeed it rocks oh Does it come with a 2.6.x kernel? and you can install a debian with IIRC 8 keystrokes and I love that it installs GRUB on there for you too :) dsmith: if you boot it via 'linux26' of which 6 are *** Morphous has joined #gnuenterprise ah *** mixi has quit IRC can you select the kernel at boot time? linux24 linux26 hurd10 ;-) *** johannesV_ has joined #gnuenterprise *** mixi has joined #gnuenterprise *** Amorphous has quit IRC *** mixi has quit IRC *** mixi has joined #gnuenterprise reinhard: That sure shut everyone up sigh yeah you guys don't even need me just a 5 minute cron job to kill svn and restart i'm tempted to go back to the source installs *** mixi has quit IRC are you running the version form testing jamest ? jamest: we also want to hear your opinion about the new .gcd format :) from* *** mixi has joined #gnuenterprise I think I slurped it from unstable ash runs unstable not testing ok we tried testing on it but had issues you know, how packages will be defined and likes *** johannesV has quit IRC sigh, i've got over 6 weeks vacation built up and desperately need to take some to try and catch up on life but when I tried to take a week off they blocked it ugh that's crap I think you deserve a vacation jamest :) i'll get it later but now's a bad time w/ courses starting up svn works for me i didn't change anything either btami: i saw the brief discussion on it last friday(?) but didn't see any details on gcd i'll try to commit gcd2sql today :) (which is part of appserver of course) *** mixi^ has joined #gnuenterprise * Sacha_away thinks maybe johannesV_ is also a machine jamest: thx, it's ok now *** mixi has quit IRC johannesV_: can you paste a little .gcd here? as a proposal any objections against a release this week? no objections :) uh, i'v forgot it, sorry :) reinhard: i got my sample running. biz-app <--> ebxml bsi <--net--> ebxml bsi <--> biz-app biz-app is custom ruby app (no real app, mockup app) so basically I am ready to redo something with gnue-appserver (if i have time) cool and i am willing to do it with gcd ;) and calculated properties and triggers reinhard: release of? jamest: common, appserver, and whatever else you think makes sense for example forms because of the fixed dropdowns jamest: Maybe doing "hot backups" instead of --dump would be better? Or maybe using --incremental? (for the svn problem) reinhard: sure, i'm cool with it *** dcmwai has quit IRC *** dhill1 has joined #gnuenterprise *** jcater has joined #gnuenterprise *** reinhard has quit IRC a commit from johannesV_ :) :) well this is a 'intermediate step' only ... gcd's really make sens if the transformation runs into a datasource directly so, this gives the chance to check if everything is available via gcd ok maybe i can add that DDL-stuff to the dbsig2-layer this week but this needs a bit more of discussion and thinking ... :) have to leave for now, maybe i'm back in the evening ... fwiw, ddl goes in gnue.common.datasources.drivers..Schema.Creation jcater: ah, i've thought there was only Introspection working well, then i've to look at this Schema.Creation ... thanks I never said there was anything in Creation but that's where we planned to put it ok, i would have placed it there too :) but as i've said, this needs a bit more of discussion ... *** reinhard has joined #gnuenterprise back :) jcater: is there something like an api or a proposal for Schema.Creation ? I don't know I don't think so, but don't remember ok jcater: stupid question.. did you commit a patch to fix the dropdown box population issue for wx ? i.e. the one that barfs when too many items loaded in it no couldn't find my notes I worked on it but couldn't figure out the fix i tried to do static definition (as i found sample in designer) and then "post load" the list the code worked, but the behavior remained the same :( iirc, it makes no difference how you populate it in our case i think we will be okay in the sense right now our screens are incomplete that is we are pulling ALL choices are you using these for the user to actually select stuff not all VALID choices or is it display only? when we narrow the list boxes down to "valid" only it will be much shorter they are selection fields (input) ok but the forms are still very raw so when i say get all accounts it will list 2500 but in reality if i say get valid accoutns for dept 200 that will be a list of 5 oh and for the record kicking your computer doesnt fix the bug either both dhill1 and i tried that patch last week and all we got was a sore toe *** mixi has joined #gnuenterprise *** mixi^ has quit IRC computers suck. *** mixi has quit IRC *** mixi has joined #gnuenterprise dhill1: yes they do gack apt-listbugs is broken for NTLM proxy server authentication i went to look at the ruby URI code * derek doesnt understand how people at the local LUG rant at how great Ruby is * derek hugs his python and gots back to building dynamic gpd generator heheh *** reinhard has quit IRC *** dcmwai has joined #gnuenterprise *** Morphous has quit IRC *** btami has quit IRC *** Amorphous has joined #gnuenterprise *** johannesV_ has quit IRC *** johannesV has joined #gnuenterprise *** wendall911 has joined #gnuenterprise *** SachaS has quit IRC *** SachaS has joined #gnuenterprise johannesV. is the sample.gcd working? i have not tried it no gcd proposal and nothing i saw in logs? all gcd work being done off line? SachaS: yes, sample.gcd is working *** johannesV has quit IRC *** dcmwai has quit IRC early night for me night *** SachaS has quit IRC *** sussman has joined #gnuenterprise mind if I list this project on the svn propaganda page? sure just don't link to our irc logs :) :) as we've ranted a bit in here never heh, yes, I'm helping dsmith in #svn with the svn problems you've been having jcater: we moved into our new building a few weeks ago http://subversion.tigris.org/project_faq.html#permissions that FAQ has the answers to your svn problems and in the alley on the way to the cars someone craps every morning it is "downtown" (read lots of homeless folks) last week someone must of cleaned up the turds we noticed and commented however someone continued to crap there so today we come to work and there is a 5 gallon buck where the turds normally are sussman: I doubt that is our issue and in black marker is written we took care of our permission issues a good while back (though, those are quite frustrating) SHIT BUCKET : Please dump in trash around corner when done our problems seem to stem from multiple heavy users at same time and next to it was a roll of toilet paper i about died (i.e., our website update script + svnadmin dump occuring at same time) jcater: dsmith and I were guessing that it happens whenever you forcibly kill 'svnserve' how so? that forcibly kill svnserve thing was added a week or so ago to try to remedy the stuck svns that hasn't been there long well, if a process has the database open, and you kill it, boom, instant wedged database. guaranteed. the two things that cause wedges are either that (a process not closing the database cleanly), or permissions problems. maybe so but that is new fwiw I think it's deadlocking wassup sussman ? wassup! jcater: the 'dump' isn't stressful, really. derek: nice story :P it's no different than a single client doing a checkout. whatever the case, make sure you run 'svnadmin' as a user which won't mess up repository perms. what's wierd is that it give lack of memory error sussman: that may be so not a permission error but our repository dumps are huge approaching a gig, iirc and the system has plenty free memory and swap at the time or maybe it was closer to 1/2 I forget but full checkouts and dumps take forever and are huge I think it all goes to crap when two of these happen at same time something else odd is that an svnadmin recover on the repository fails with an error telling us to run svnadmin recover aha! yah, you need to run catastrophic recovery db_recover -c -ve -h repos/db/ which replays all the logs out of memory error, you say? yes you might want to try increasing the maximum number of locks in your repos/db/DB_CONFIG file after doing that, run recovery. see if that fixes the problem. ok, hang on a second and I'll give it a show shot that command we're hearing other bug reports about this... it seems to be an elusive scalability bug that we're tracking. see if that workaround helps. in our case our repository is at /var/svn/gnue db_recover -c -ve -h /var/svn/gnue/db right? hmm cp -R that directory first :) :) was going to tar zcf it is it live? if so, you should do a hotcopy cp -R may result in a copy that's not recoverable. that's the whole point of "hotcopy" ok, how do i do that? um, 'svnadmin help hotcopy' you guys really ought to read chapter 5, if you're adminning a repository. got it db adminning ain't so simple. :-) yeah i read it before i started using it I think that's one of the disadvantage + advantages of svn the fact that it's db-based but i've not had this kind of trouble with an svn install it's always "just worked" well, svn 1.1 has a plain-filesystem repository option now. oooh it may make life easier for small setups let be corporations deal with db adminning let big corps, I mean hmm that's cool I suppose w/plain-filesystem, finally anonymous checkouts won't have to have write-permissions on the directories? sussman: Ya, I've seen some people get scared when they hear "binary databasse" correct with a plain-filesystem, you could run a repository off a CD-ROM svn 1.1-beta will be released a couple of weeks, I'm guessing i don't mind the db thing other than the write access anon svn is down, ssh users please stay out for a bit * derek goes to run that killsvn script that spawns an svn check out on all 13,000 pcs at the count at the same time..... no need derek, just your being here is enough jamest: And remove that killall in the backup script. db_recover: PANIC: fatal region error detected; run recovery derek: see jamest: yummy... i love soem panic in the mornign * derek checks his fingers.. they appear to not be typing properly today must be time for lunch derek: yes, please leave or at least step away from the computer um i'm trying to fix things here and you're getting your karma all over things derek: please keep your karma to yourself *** kfogel has joined #gnuenterprise i'm running the db4.2_recover as the debs seem to require 4.2 db4.2_recover -c -ve -h /var/svn/gnue/db was the exact command *** btami has joined #gnuenterprise hey btami please hold off commiting anything for a bit ok, i have nothing to commit now :) jamest: what's the output of your recover command? which one? svnadmin or the db4.2_recover? (Or were you just indicating that you are running it, not saying there was a problem?) db4.2_recover there is a problem What is the problem? here is the ending db_recover: PANIC: fatal region error detected; run recovery db_recover: DB_ENV->open: DB_RUNRECOVERY: Fatal error, run database recovery ash:/var/svn# erm a bunch of that first line You are running as the same user who owns all the db files, right? all the way back thru my buffer (and owns all the log* files, etc) running as root That's the answer to a different question :-). the owner of the files change ??? but the group stays the same ash:/var/svn/gnue/db# ls -al total 111428 drwxrwsr-x 2 root gnue 4096 Jun 7 14:54 . drwxrwsr-x 7 root gnue 4096 Mar 19 22:46 .. -rwxrwsr-x 1 root gnue 1738 Mar 11 12:35 DB_CONFIG -rwxrwxr-x 1 root gnue 7843840 Jun 7 14:53 changes What is ls -l of your db/ subdir? -rwxrwxr-x 1 root gnue 36864 Jun 7 14:53 copies Thanx -rw-rwSr-- 1 johannes gnue 1048531 Jun 4 15:03 log.0000000818 -rw-rwSr-- 1 jcater gnue 813584 Jun 7 14:53 log.0000000819 -rwxrwxr-x 1 root gnue 9711616 Jun 7 14:53 nodes -rwxrwxr-x 1 root gnue 9216000 Jun 7 14:53 representations -rwxrwxr-x 1 root gnue 139264 Jun 7 14:53 revisions -rwxrwxr-x 1 root gnue 83030016 Jun 7 14:53 strings -rwxrwxr-x 1 root gnue 2093056 Jun 7 14:53 transactions -rwxrwxr-x 1 root gnue 8192 Jun 7 14:53 uuids svn, svnserve, and svnadmin are wrappered via You've read http://svnbook.red-bean.com/svnbook/ch06s05.html, right? ash:/var/svn/gnue/db# more /usr/bin/svnadmin umask 002 /usr/bin/svnadmin.real "$@" if [ "$LOGNAME" = "root" ] then chmod g+sw /var/svn/* fi ash:/var/svn/gnue/db# more /usr/bin/svn umask 002 /usr/bin/svn.real "$@" ash:/var/svn/gnue/db# more /usr/bin/svnserve umask 002 /usr/bin/svnserve.real "$@" This method could be made to work, with sufficient care, but I should ask, why do you want to do things this way? Why not use a standard setup where the same user/group owns everything, and the processes that access the repository all run as that user? we do checkouts via svn+ssh *** fitz has joined #gnuenterprise this seemed to be closest match to what we had before bbiab All the above is on a hot copy, not on the real repos, is that the case? *** reinhard has joined #gnuenterprise i made a hot copy but was trying this on the real deal as we can't get much worse :) i have anon svn off and requested the people w/ ssh to stay out for now fun w/ svn... easiest thing is to use apache2 ;P (mayeb not that secure though unless you use https) bah lots of overhead for little gain humbug * jcater looked at apache2 at work was easy for me chillywilly: we're on a limited resource machines machien machine that's being used for more than originally intended I don't think apache2 is that big it's not that it's that big I didn't see the gain from using apache2, though it that ash already has a fair bit of crap running on it ref: straw, camels back so, have you shut down all access to the repository? heh start with that. all that I can then 'cp -R' it. and play with the copy. Okay, so catastrophic (-c) recovery isn't working. What does 'df' on your machine say? apache2 might more scalable than svnserver (no offense) i can't kill ssh as we've got no physically access to the machine if things go wrong I've lost track as to which copy is currently more hosed. Are you coming through an NFS mount or anything? no nfs mounts okay yah, it's on /var hmmmm Can you tar up that baby and let us download it? ash:/var/svn/gnue/db# df Filesystem 1K-blocks Used Available Use% Mounted on /dev/hda1 37487248 11037720 24545244 32% / tmpfs 112044 0 112044 0% /dev/shm i can give you access to the machine itself if you like * kfogel wonders whether having shm in tmpfs could possibly have anything to do with it "SVN Cracks GNUe, Read all about it" jamest: sure, that would be great. But you probably don't want to give me root, so I'm not sure that will help. Unless you change repos perms. s/perms/ownership/ What's in DB_CONFIG, by the way? huh??? gnue@gnu.org gives "no such user"?? don't you mean gnue.org? if necessary i can leave and run to the building this machine it is in so that i can give it some real loving probably did not set it back up after savannah was cracked karma delight so to speak derek: don't you get near it reinhard: I was noticing that this weekend I can't log into the mailing lists don't they run mail on the same box? to approve messages weird jcater: i mean you *posted* a message to that list just a few minutes ago or maybe an hour or two * kfogel gently pings for DB_CONFIG contents again, no hurry, just askin' set_lk_max_locks 2000 set_lk_max_lockers 2000 set_lk_max_objects 2000 set_lg_bsize 262144 set_lg_max 1048576 set_flags DB_LOG_AUTOREMOVE appear to be all the "non" commented lines sussman: kfogel@ash:/var/svn/gnue/db$ pwd /var/svn/gnue/db kfogel@ash:/var/svn/gnue/db$ ls -l total 112240 -rwxrwsr-x 1 root gnue 1738 Mar 11 12:35 DB_CONFIG -rw-rwSr-- 1 www-data gnue 8192 Jun 7 15:01 __db.001 *** holycow has joined #gnuenterprise -rw-rwSr-- 1 www-data gnue 270336 Jun 7 15:01 __db.002 -rw-rwSr-- 1 www-data gnue 327680 Jun 7 15:01 __db.003 -rw-rwSr-- 1 www-data gnue 737280 Jun 7 15:01 __db.004 -rw-rwSr-- 1 www-data gnue 16384 Jun 7 15:01 __db.005 hey all -rwxrwsr-x 1 root gnue 7843840 Jun 7 15:08 changes -rwxrwsr-x 1 root gnue 36864 Jun 7 15:08 copies -rw-rwSr-- 1 johannes gnue 1048531 Jun 4 15:03 log.0000000818 -rw-rwSr-- 1 jcater gnue 826142 Jun 7 15:08 log.0000000819 -rwxrwsr-x 1 root gnue 9711616 Jun 7 15:08 nodes -rwxrwsr-x 1 root gnue 9216000 Jun 7 15:08 representations -rwxrwsr-x 1 root gnue 139264 Jun 7 15:08 revisions -rwxrwsr-x 1 root gnue 83030016 Jun 7 15:08 strings -rwxrwsr-x 1 root gnue 2093056 Jun 7 15:08 transactions -rwxrwsr-x 1 root gnue 8192 Jun 7 15:08 uuids kfogel@ash:/var/svn/gnue/db$ sussman: notice the www-data lines I think that's webcvs. jamest: is webcvs wrapped the same way as all the svn+ssh stuff? svn is back up i presume? holycow: no, working on it derek: not sure if you know this one but grep -v ^# DB_CONFIG holycow: no the svn hackers are here helping ;P k. :/ * holycow ^5's all the hackers :) we'll owe them a beer or soda or something How about a Free enterprise system? :-) i'll send a 6 pack of jones vanilla soda :) official gnue goat sex? lol lol, you'd get the soda quicker I imagine :) * kfogel raises eyebrows kfogel: I' hi kfogel m pretty sure it is it uses the svn commands, iirc we've never had it's use lock up svn during the day the DB_CONFIG settings look absolutely standard jcater: our day anyway :) "gnue" is our master group and those www-data's are part of "gnue" group, with group read/write jamest: I'm positive this has to do with either our svn dumps or our nightly checkout as those are always left hanging and derek I don't think it has any thing to do with reinhard, et al, being a lot more productive than us :) i.e., while we're asleep lol @ jc :) /msg jamest though that could be the case i was thinking if we helped them debug this problem maybe the would give us their software for free on the next release ;) i just rolled gnue.backup over to the gnue group *** kilo has joined #gnuenterprise jamest: I'll try a recovery there, just to reproduce the prob kfogel slightly off topic... do you have anything to do with the new svn book? derek, reverse psychology? ehehe i like it :) * derek tried to get someone here to procure it the other day, but it wasnt available (yes i have an online copy) db4.2_recover is the command, right? it appears to be procure it, as in, the book? derek: no, Ben Collins-Sussman, Brian Fitzpatrick, and C. Michael Pilato (all my coworkers) are the authors. * fitz waves I wrote the foreword, though. :-) as that is what the debian subversion package requires cool hi fitz :) greetings. well Maricopa County is using svn and wants to buy the book :) * kfogel whistles i think we have it "ordered", but its not shipping derek: maybe you could just promise not to use their software for a while derek: *everyone* in Maricopa county should have a copy!!! Why buy the book, when you can buy Fitz instead? derek: send me a check for $50 * fitz slaps a pricetag on his forehead and I'll get you a hardcopy :) rofl fitz... just $0.99? fitz: how much you going for today? bargain!!!! ill take two! where is Maricopa county? and are you willing to move to the desert and how many users? sussman.... Arizona its in arizona someone set up subversion in county offices? phoenix, scottsdale, glendale, gilbert, mesa etc are cities there currently only the development team for the new budget system is using svn group of 4 or 5 if it works, why not cool too hear :) i hope they are dev'ing on gnue? well we are dumb enough to use gnue *mmhhmm*? jamest: I'm looking at http://www.webservertalk.com/message203777.html right now, Keith Bostic's reply has some useful bits I think :) so why not svn ;) derek, have you checked out this --> http://www.osafoundation.org/? dont get me started heh quick answer yes the former lotus head is developing something similar again in python and both them and tony stanco's group drvie me mad i'm gonna introduce him to gnue :) the git has nothign about gnue on his site ah really? hehehe kfogel: odd i don't think anyone's been removing log files perhaps i should then just shush and let them dotheir own thing holycow : no not really i was thinking of a different group osa is out of austin? right i think i have talked to them a bit jamest: autoremoval is (see DB_CONFIG). But, good news: woo. if they are the ones writing chandler In the backup, I just did a non-catastropic recovery (i.e., eliminate the -c flag) and got chandler yes... i just ran it... nothin special kfogel@ash:/var/svn/gnue.backup/db$ db4.2_recover -ve db_recover: Finding last valid log LSN: file: 819 offset 810950 db_recover: Recovery starting from [819][810098] db_recover: Recovery complete at Mon Jun 7 15:40:16 2004 db_recover: Maximum transaction ID 8000c98f Recovery checkpoint [819][810950] kfogel@ash:/var/svn/gnue.backup/db$ I'm going to try it in the real db now. derek: make sure Maricopa employees buy lots of harcopy O'Reilly subversion books. :-) kfogel: cool i actually talked to texas university in austin when going to eval new bud system we need to repay our royalty loans. :-) jamest: okay, try accessing svn now (but stay tuned even if it works) and we talked a lot about chandler as they were putting monies into it (iirc) lol, permissions got fried ah really? i presume gnue came up? sussman : unfortunately i loathe o reilly :( i looked at some of what they seem to be doing, frankly i think they could do better may i ask why you didnt go with a publisher like no starch press or gnu press? sussman, no worries, i'm buying a copy this week :) jamest: oops, sorry derek: why loathe o'reilly? ideology of its owner kfogel: we're used to that which is why we wrappered svnadmin and friends :) btw: i still would recommend the book jamest: sorry. You must have a script to fix perms somewhere... it's fixed but any ideas on why? why it's happening as it's daily * kfogel wonders mildly about Tim O'Reilly's ideology, but would rather concentrate on prob with jamest for now, so won't pursue it further atm jamest: yes, some ideas, one sec (researching) derek, just out of curiosity, your rough take on chandler is ... ? it would seem we could do all that and more here jamest: I worry that it might be this: http://subversion.tigris.org/issues/show_bug.cgi?id=1817 i havent looked too hard at chandler And even if that's not the cause, it's *still* something to pay attention to. ah cool, was just curious. nm then holycow: the svn book won't be in bookstores until June 30 If you guys have a regular backup schedule, using hotcopy or some other similar method (perhaps one based on contrib/server-side/backup-recipe.sh), then you probably want to turn off auto-removal of logfiles in your repostories. not out yet? ah! k. jamest: Okay, so high-level, no-nuances view here: If -vce isn't working, try -ve (in db4.2_recover, I mean) I know you tried -ve first and it didn't work. Not sure what to say about that :-). I just got lucky somehow. The reason is that the -c needs all logfiles from time of last checkpoint. we've been svnadmin dump'ing the data, not hotcopying should we be hotcopying? sorry, I mean "all logfiles", forget the part about from time of last checkpoint. no, due to http://subversion.tigris.org/issues/show_bug.cgi?id=1817 and http://subversion.tigris.org/issues/show_bug.cgi?id=1818 ... I would say don't use hotcopy, unless you turn off autoremoval of logfiles. Which you might want to do. Basically, the problem is that SVN's backup tools are not mature yet. Which sucks. That one I just mentioned, contrib/server-side/backup-recipe.sh in the Subversion distribution, is known to work under high load and not be a problem, *if* you don't have autoremoval of logfiles turned on in your DB_CONFIG. However, that script is not a tool per se, it's just an example of how to write a full+incrementals backup recipe. Anyway, the backup stuff might all be a red herring, it could be that the cause of this problem is something else entirely. Unfortunately, svn+ssh is prone to this sort of thing, because it's really hard to make absolutely sure that *every* accessor of the repository is behaving well. they have a script which 'kills' every svnserve process that's a high risk of wedgitude I know your box is underpowered, but if you can possibly do https:// instead, or just settle for normal svn://, that would be better. that can come out Since this is a free project, do you really need the ssh:// part? oh! as the kill was an attempt to clean out the wedge yah, does it have to be encrypted? :-) sussman: yep, sure is . we were wedging before "before" means when? but if we killed all instances of svn* then things worked again jamest: btw, is everything working now? ugh as we're have the dump hanging along w/ some checkouts Failures wouldn't be reliable. Sometimes you'll have them, sometimes not. it's svn up ing oops, I wrote "ssh://" where I meant "+ssh", hope I didn't confuse anyone. jamest: cool (about the svn up) kfogel: i followed :) good heck, regular svn:// would use less resources anyway, so that fits in with the low profile you're aiming for... we also want secure developer logins how secure? no passwords in plaintext Oh, you mean, shell login? Of course, you can still have that, it's totally independent of svn. yeah, we have developers log into ash via ssh Just tell people not to use the same password! :-) this seemed the easiest way to get them svn access are you kidding? svn+ssh:// is a permissions nightmare no you saw the checklist, I assume the proximity of "svn+ssh" and "easiest" in the same conversation does leave me wondering :-) getting the right group ownership, sgid stuff, umask stuff, etc. we came from savannah it's a nightmare. cvs+ssh if you run just 'svnserve -d', you're done. but once set up, it's not a big deal a nightmare, I'm telling you sussman: tell him what a nightmare it is one svnserve daemon running as *one* user who wholly owns the repository. for the developers this was a small change yes, but listen to what I'm saying: one svnserve daemon running as *one* user who wholly owns the repository. not N users all trying to access the database as themselves. get it? yes sussman: you're telling him what to do, not what not to do, right? you have a single point of breakin :) of course sussman: Yeah, but how do you handles the passwords for that? Isn't just a text file? much, much less risk of permissions being munged. yes, it is. dsmith: yes, but big deal. but the password isn't passed over the network. oh, right, good point it's CRAM-MD5. The password can't be sniffed safe enough. exactly. ah so the safest path is to create an 'svn' user, and run 'svnserve -d' as that user. and make the repository owned by 'svn' then have users all use svn:// * kfogel thinks maybe the svn:// documentation should make this clearer, as a lot of people have shied away from svnserve under the misconception that it's on-the-wire security is as bad as CVS's. CVS pserver, I mean yah. s/it's/its/ i know that's exactly what i was thinking and whenever you run 'svnadmin dump' or anything, run it as user 'svn' as well. to play devil's advocate though... we know the risks with using ssh we don't with svnserve jcater: I think after today you know the risks of svn+ssh too :-) it's a new daemon running on our precious baby we run svnserve already but it's not listening on port 3690. kfogel: fwiw, both jamest and I also use svn+ssh at work as i had to when we added anon svn using 'svnserve -d' means opening port 3690. well, they can use some custom port, right? and speaking a custom protocol. we run svnserve via inetd now, fwiw i have svnserve down atm we we started messing w/ it today svnserve -i I mean, both client and server allow you to specify something other than the default, I believe. which i shut off today Anyway... jamest: I guess you're set -- you have all the information, the decisions are of course yours to make. i'm ok w/ trying svnserve, i assume you can set > 1 user/passwd for it or is it one only? I'll go back to work now, but feel free to ping if anything comes up. more than 1, of cuorse danke guys :) to where do we send the soda? :D I think the book covers it, but ping if problems. thanks guys, and thanks for the passwd info as I swore it was plaintext Good luck with GNUe, it looks exciting! the passwd file on disk is plaintext, that's true. so make it chmod 700 but no plaintext passwords are passed over the net. sussman: Are there any tools to edit that file, beyond a text editor? that's all you need. :-) someday we may write something similar to htpasswd. but not yet. * dsmith is thinking of the poor windoze users dsmith: an admin interface would be great, want to write one? :) dsmith: ? they don't have editors? only warezd ones lol :) Not on the unix boxes, corewrite only runs on m$. kfogel: Any chance of hooking svnserve into pam? dsmith: it's been discussed, Greg Hudson the svnserve author has been wrestling with that and related issues. Post to users@subversion.tigris.org to find out details (I don't know them). kfogel: Not "when are you guys gonnaa do it", but do you think it's a good/bad idea? kfogel: ok Well, good idea in principle of course, but my knowledge of PAM is sketchy, so I don't really have an informed opinion. it's funny. people are really opinionated about the issue of local accounts. the Apache Software Foundation is switching to apache+svn because they *don't* want to give out unix accounts for cvs. meanwhile, others *insist* on PAM integration, even in apache. or svnserve. sussman: but the beauty of PAM integration is not in the local accounts its in the abstraction ah, yes, say I have an LDAP-based authentication system set up for everything else kfogel: One thing that was tripping me up about svn was that https: svn: and svn+ssh: all used *different* pwasswords, and different methods of maintianing them. but most people say, "give us PAM, because we want to use existing system accounts!" then svnserve could tie into those LDAP accounts too dsmith: yes, because they're all different authentication systems. apache, svn, and ssh. 3 separate worlds. sussman: Exactly you're in a maze of twisty options, each subtly different. heh to be fair, i havent read the back log entirely two issues with just plain svn 1. for people like us government users behind highly restrictive firewalls... its painful 2. the svn printed documentation on admin functions was severely lacking... imho... derek: that's exactly why we have the http:// option... (for 1, I mean) there was a ton of confusion as to what the options were and the impact admin functions? you have 4 sources granted i think a lot of this is because some of it is different than cvs so trying to forget what you know about cvs can be a pain :) chapter 5, chapter 9, 'svnadmin help' 3 sources, I mean im talking chapter 5 "Fear, surprise, and an almost fanatical devotion to the Pope." nobody expects the spanish inquisistion! Ni! For the record, that was unrehearsed :-). sorry it is chapter 6 not 5 that was not all that clear kfogel, haha thats brilliant my brain hurts, too much learning for one day :) particulary it seemed to appear that you had to already know a bit about apache + mod_dav if you wanted to go the mod_dav route though perhaps that is not fair give me a break tis true. it's easy the svn book deliberately doesn't go into apache tutorials. there are whole books written on apache. as it looked a lot more complex to setup so i just glossed over it apache is the shiznit we had to draw a line. and decided for the other way was easier I can set it up for you derek :) * derek is not an apache expert and am on deathlines for projects so i needed something quick *** reinhard_ has joined #gnuenterprise ah, here comes trouble that and i think when james and jason were steering away from it, made tme decide to back off sussman : ie you are correct... you have to just tell the config parts pertinent to svn yah. or your book would be 1000 pages exactly. come by #svn anytime, guys. l8r just set up a vhost and throw their config directives in it *** sussman has left #gnuenterprise I used the ones right outta the book bye, thx again good luck guys! *** fitz has left #gnuenterprise well and you need to setup htaccess file(s) and that's it is it safe to svn again, or are you guys still performing modifications? but at the same time, the online books say that the htaccess method isn't recommended as it isn't very secure do it over https then that only hides the password nothing more from a risk management standpoint so? one of the reasons we went with svn+ssh for gnue https only "hides" your CC information too ;) holycow: I believe it's safe, but ask jamest to make sure first my cc information isn't a static entry to the system either but your ssh username and login are dynamic? chillywilly: does your bank use htacces for getting you in? probably uses a database and gathers the info on a https form htf that is any more secure than htaccess is beyond me *** fitz has joined #gnuenterprise jcater: https hides everything, not just password htaccess with https that is it's total encryption end-to-end yes I never denied that Were you talking about https:// repository access, or something else? encryption is only one piece of the security puzzle, though I was talkikng about that kfogel kfogel: I'm just arguing with chillywilly jcater: and making an ass of yourself :) Hmm, I must have misunderstood something due to lack of context, then, sorry. chillywilly: not as much as you think no serious corp would use htaccess for authentication I understand your point but who in their right mind uded system usernames and passwords for htaccess uses* *** reinhard has quit IRC *** sussman has joined #gnuenterprise system usernnames/password are a *really* bad idea. (for http) jcater: but if is encrypted it won't matter...I only use plain htaccess to keep google away (and maybe https too) dsmith: yeah, just give me a pint of blood per login and we'll be happy :) and i'll turn anon svn access back on now The problem is it's too easy to brute force attack it. ? anyway, what everyone really wants to use is digest not plain ol' http auth but browsers don't support it jamest, let us know when its safe to download some :) holycow: give it a rip danke sussman, kfogel, fitz: thank you btw, i was called away from desk before I could thank you the first time np what URL should I checkout? svn:// jamest: you're welcome sussman: for our stuff? yah, just curious. svn://www.gnuenterprise.org/var/svn/gnue/trunk gnue jamest: "For another, Web "logins" don't have the same security controls built in as have been added to most operating systems in the last few years; over the Web, an intruder can keep hammering away at a username with password after password without the system taking any defensive measures. Most operating systems nowadays will enter a paranoid mode and at least ignore login attempts for a while after some number of failu res; all mod_auth will do is record a message in the Apache error log." svn co ........ :) wow, you guys have a POS system? dsmith: ah, i thought you were dissing username/password in general PoS too sussman: yeah, the whole thing is a pos er, wait nice! HEHEH actually the PoS was used by jcater in his bookstore I knew there had to be a Free PoS system out there somewhere, cool. *** fitz has left #gnuenterprise i know he had it working w/ a cashdrawer dsmith: stop diostracting me ;) distracting...*sigh* but I'm not sure how much made it into our repository WHat? jamest, rofl :) I'm surprised you guys aren't using arch instead of svn, this being a GNU project and all. sussman, gnue is pretty much THE SHEIT, i've looked everwyere for creating sensible erp type stuff sussman: we also code in python, not exactly the GNU standard :) * sussman hearts python nice, got the whole trunk in 2 minutes. jamest: everything but my database backend is in svn it operated a two-color register tape + a cash drawer + scanner I wanted a pole display are there standard hardware protocols for these pieces of equipment? but never found one cheap enough on ebay sussman: yes and no about as standard as printer protocols sussman: Thanks again for helping these guys i.e., epson has a ESC/POS and a lot of receipt printers emulate that language for every piece of equipment, there are probably 3-4 "standard" ways it could work dsmith-away: it's cool to see projects using svn. gotta help the cause. sussman, the nice thing about gnue as i'm sure you have gathered already is that as a forms based environment, the server makes it easy to publish anywhere, from what i'm told one just needs to write a driver to publish to said platform/target wow with other things out there that have similar products in place, they are tied into a particular delivery vehicle that's the promise of gnue-forms however atm the best front ends are the WX one and win32 one *nod* of course it is a fairly young project with limited resources, would be nice to start seeing more interest in the idea as for backends we basically attempt to support anything python's DB API provides a driver for on the RDBMS side gtk2 isnt all that bad young? * jamest blinks ehehe *erm* sorry :) right hehe forgot the history bits, its young in terms of my personal radar, lets just say that :D it's depressing how slow we are jamest: it's young in the same sense that you are young at heart, at least * jcater hides ah lol :) thanks there is something to be said for slow yeah but usually it's not nice and knowing the code like I do, it's just as twisted and bitter on the inside as I am too i'm a designer by nature and let me tell ya, the best stuff takes time. theres this funny thing that after the final edit there is a process of osmosis to get it just right i look around at things like zope and see fantastic work, but they grew so fast they have adopted certain peculiarities into their projects as a result thats neither good nor bad, i just see everyone talking about something like gnue, but only you guys are really building it sussman: if you're into python then you might find gnue-common interesting oh? holycow: zope is switch to svn too. :-) eh really? nice :) it's the core of all the gnue apps and is really a general purpose python based application framework if you look in gnue-common/doc there will be a Developers-Guide.sxw that starts to explain how you can use common mmmmm while it starts off using the entire framework a great deal of common can be used in any application night all *** reinhard_ has quit IRC night I love GNUe chillywilly: you still using it in your stuff? no... i love the concept, i now need to start learning it ;P our docs do suck though as that guide isn't complete I was using python-dbo.sf.net which there's no stinking docs about either jamest: I have two friends running a bookstore in Krakow, Poland entirely on free software, if you get good docs, they'd probably be *very* interested in gnue. (Heh, now, I'm sure two people in a bookstore in Poland are all the motivation you need, right? :-) ) haha lol http://www.massolit.com/, btw i think the only thing that is needed on the site is a page where people can upload their projects anonymously, forum style jamest: my "config" system consists of a python module with a dict in it ;) share ideas back and forth, gnue needs a library of experiments imho kfogel jcater runs a bookstore on gnue i.e. gnue-pos in cvs is built to run a bookstore well, originally a bookstore then we started a 9,500 sqft retail store antiques and gifts sorry to be such a newb, i get a 'no such file or directory' on svn://www.gnuenterprise.org/var/svn/gnue/trunk: No such file or directory. i thought the svn command i tried at home was different from this? jcater, v cool :) holycow: you need "svn co " in front of that ah! i thought something was amiss, danke i just gave sussman the end part :) ah right, svn 'c'heck'o'ut, i get it :) so jamest, you run your own store in terms of a career? i mean as opposed to a programmer per se? Krakow, Krakow... oooh, I did got drunk there....... (too)..... i'm also using gnue common inside a mod_python system and I'm attempting to write an RPG with common (actually i have a bit of it done but no free time to work on it ) holycow: jcater did not me RPG? role playing game pet project jamest rewrites quake3 in gnue,,, using curses of course holycow: they're my wife's stores ah! right right, neat. usually these kinds of projects involve purely coders who are very good at abstract stuff, and lack a real world use case :) kilo: nah, not a first person shooter guy jamest: yah, I'm an RPGer too. jcater, :) v awesome sussman: cool jamest: I'm an openrpg developer and have a weekly group dbl cool url? uebercool :) oh, um www.openrpg.com http://www.openrpg.com/index.php?page=features jamest: oh, how nice it would be: from gnue.common import bloodshed, torture, rocket.launcher sussman, neat idea a bit similar to gnue :) it's all wxpython, yah. but really, subversion is my day job. :-) kilo: those are stored in the gnue.forms and gnue.reports modules rotfl night all *** btami has quit IRC okay, are there any instructions on how to run gnue from the cvs download dir? is that still possible? yes cd gnue-common ./setup-cvs.py will work on a unix box i don't know the windows way nite *** jamest has left #gnuenterprise check the wiki for windows install instructions i'm on debian :) ok then, check the wiki for debian install instructions 8-))) hehehe *nod* i'm installing as we speak, i just couldn't find the instructions for cvs local install, cool *** sussman has quit IRC *** kilo has quit IRC wiki? gnue has a wiki? mais uoi oui even where? * chillywilly looks www.gnuenterprise.org look on left sidebar *** wendall911 has quit IRC lol there she be *** wendall911 has joined #gnuenterprise rofl best quote ever re: ms merging with sap discussion Microsoft would merge with SAP the same way I merge with a cheeseburger. rofl :D look on left sidebar err, wrong window *** dhill1 has quit IRC * kfogel is away: headin' home good job jcater who do all the svn people work for? four of them work for CollabNet, inc (www.collab.net). The other fifty six work for their respective employers, if any :-). how does collabnet make money? Selling collaboration software and hosting for same -- see www.collab.net for details. It's like "distributed, collaborative project in a box" yea I am there Sun uses them apparently thats cool Press a button, supply some information, and you're off and running: mailing lists, web pages, revision control, bug tracker, file upload/download area, it's all there, and you don't have to reinvent the wheel. Yes, it's quite nice. I see great to see in action what everyone in the open source community has been preaching, it's about the services and added value so they do any free hosting? The big win is unified user/group/role management, which is especially nice for corporate customers. or only of their ow products like svn ;) own* chillywilly: I think they donate hosting to the openoffice.org project I see well, tigris.org is free hosting, though note that the version of the CollabNet software it runs is always a bit behind. if you look on their website, it has "powered by collabnet" # Sun Microsystems: * java.net * Jini Network Technology * Grid Engine * SunSource.net * Project JXTA Yeah. OpenOffice.org is also running CollabNet. okay, really gotta go home, see y'all later l8r yeah, love to see someone take the open source thing and make money with it bye kfogel holycow: well, the software isn't all open source, so not sure how well that fits the model you had in mind (subversion is open source, of course) * kfogel is strangely addicted to irc today, puts on shoes, *really* goes home :-) lol Tim O'Reilly is no the board on* damnit I can't type I should eat hey ppl what's your favorite window manager (i.e., non-desktop env) I wanna install something less resources intensive on my old powerbook *** wendall911 has quit IRC *** Amorphous has quit IRC *** Vee2d2 has quit IRC *** Vee has quit IRC *** havoc has quit IRC *** bluesbaron_ has joined #gnuenterprise *** wendall911 has joined #gnuenterprise *** Amorphous has joined #gnuenterprise *** Vee2d2 has joined #gnuenterprise *** Vee has joined #gnuenterprise *** havoc has joined #gnuenterprise *** bluesbaron_ has left #gnuenterprise *** ajmitch has joined #gnuenterprise *** jcater has quit IRC *** chillywilly has quit IRC *** chillywilly has joined #gnuenterprise *** wendall911 has quit IRC *** mixi^ has joined #gnuenterprise *** mixi has quit IRC *** jcater has joined #gnuenterprise *** dcmwai has joined #gnuenterprise *** holycow has quit IRC *** holycow has joined #gnuenterprise *wave* *clap* *** dsmith has joined #gnuenterprise k. downloaded from cvs installed all the debian packages setup postgresql as shown: http://www.gnuenterprise.org/cgi-bin/twiki/view/Main/SettingUpPostgreSQLToTestWithGNUeAppserverSamples ran setup ./setup-cvs.py, installed without generating errors k. how do i now startup the application server and say for example forms designer? I haven't played with appserver yet, so I don't know danke. sacha will know :) gnue-appserver gnue-designer :) those ought to be the script names, for a start feel free to file packaging bugs hmm wait are you running gnue packages from cvs, or debian? :) cvs AND debian :) lol what i mean is i'm not running any of the debian packages i cvs'd the latest down to a gnue dir then followed the directions as outlined above as best as i could *nod* gnue-appserver would work if i had used debian packages ok oh check it out someone is asking for commercial support for gnue on the mailing list we gotta get our act together, some of you devs may be able to make some extra cash :) commercial support? surely not... someone would be nuts to do a support contract for gnue right now not kidding :) i mean not kidding about someone asking *nod* *** jcater has quit IRC