IRC logs of #tryton for Saturday, 2010-11-06 #tryton log beginning Sat Nov 6 00:26:24 CET 2010
-!- irclog( has joined #tryton10:54
cedkudono: I'm wondering if we are not subject to the 1.6.4 hgweb memory leak11:04
udonocedk: but it seems to be fixed in 1.6.411:09
cedkudono: and we run 1.511:09
udonocedk: Ic11:09
cedkudono: but I don't find the correction11:10
cedkI suspect hgweb because we need to reboot often the server because of "out of memory" almost since I setup hgweb as a fcgi11:14
cedkudono: I will try it thx11:16
cedkudono: but it seems it is on client side11:17
udonocedk: ... otherwise you may use a cron, that kills the process all 6 hours11:20
cedkudono: it is what I'm looking for :-)11:21
udonocedk: but this has the issue if someone is pulling in the moment you kill his process, a corrupted mrepos remains.11:21
cedkudono: hg will stop with an error message I guess11:22
cedkudono: I just test it and lighttp wait to restore the fcgi11:23
cedkso there is only a slowdown in cloning but every thing goes fine11:24
udonocedk: so it sounds perfect.11:26
cedkudono: I would like to be sure it is this process that is eating all the memory11:27
udonocedk: so just run some kind of 'ps ux >> ps.log' with a cron first for some days to identify the fat process? Or maybe there is a way to quota the space for user processes?11:36
-!- evernichon( has joined #tryton11:39
cedkudono: I started atop11:42
yangooncedk: shall I put the links for hgnested on 1.8?11:50
cedkyangoon: don't understand11:50
yangooncedk: they are not yet made made, hg nclone for trytond1.8 doesn't worjk11:51
cedkyangoon: they are in trytond/11:52
yangooncedk: ok, missing in neso11:53
cedkyangoon: ok11:54
cedkudono: I think we must change tycho, it has only 235M of ram11:57
cedkudono: plato got 1G11:58
cedkand it is the same kind of host buy to ovh11:58
cedkbut at different time11:58
udono235M is not a lot11:59
udonoACTION away bbl11:59
cedkudono: in fact, I think we made a great challenge to get all running with this amount :-)12:00
udonocedk: :-) but we did not announce version 1.8 at all12:01
cedkudono: I still need to fix issue1748 and issue174712:03
cedkudono: I guess we will publish the announce monday12:03
-!- chrue( has joined #tryton12:03
cedkbut issue1748 is very strange12:03
cedkI think it is linked to the fact that sqlite reuse the same id when a record has been deleted12:04
cedkudono: for the same price, we can have the same host with 2G12:06
yangooncedk: same CPU or better?12:08
cedkyangoon: almost the same12:10
yangoonseems to me that  (apart from the memory leak) CPU is more the bottleneck12:10
yangoonbut 2G surly won't hurt:)12:11
cedkyangoon: I think it just fit, it is arround 0.8 of load average12:12
cedkwe can have a 4core with almost the same power for each core and 4 G of ram12:13
cedkfor 4x the price of the 1core12:13
cedkbut I think we could split the services on 2 simple hosts12:14
-!- ikks(~ikks@ has joined #tryton12:30
-!- FWiesing( has left #tryton13:04
-!- enlightx( has joined #tryton13:14
-!- heg( has joined #tryton13:17
-!- tony_(~tony@ has joined #tryton13:24
-!- evernichon( has left #tryton15:51
-!- tony_(~tony@ has joined #tryton16:10
-!- lem0na(~lem0na@ has joined #tryton16:33
-!- enlightx( has joined #tryton16:34
-!- pheller( has joined #tryton17:32
phellercedk: testing sqlite patch now17:33
cedkpheller: ok17:33
cedkpheller: we should also include the patch for issue1747 in neso build17:34
phellercedk: ok, is there a patch on codereview to try?17:35
cedkpheller: not yet but it is very simple as it is just a typo17:35
phellercedk: ok, if you can make the patch, I will test it right now...17:36
phellercedk: and if you apply this, the patch, and the sqlite patch, I can make all the mac packages in the next hour or so17:36
cedkpheller: I will wait until tomorrow before pushing sqlite patch17:37
phellercedk: ok, I can make the packages tomorrow too17:37
phellercedk: we should add a setup wizard action to create financial periods for 2.0....17:38
cedkpheller: why not17:39
phellercedk: I'll open an enhancement for this.  we should maybe define some default sequences too.17:40
phellercedk: ... so it is easier for someone who is trying tryton to get up and running17:40
cedkpheller: normally there is default sequence17:40
phellercedk: ... for a fiscal year?17:40
phellercedk: every time I add a fiscal year, all the sequences are empty17:40
cedkpheller: no because there is  sequences per fiscal year17:41
phellercedk: right, maybe this could be a part of the wizard; it could offer to create sequences with a default name, like: "FY2010_post_move_seq", etc17:41
phellercedk: the user could change them....17:42
cedkpheller: yes it could17:43
cedkpheller: perhaps the better will be to have a sequence generator for fiscalyear17:43
phellercedk: yes, I think so.  if the integrator for a real installation does not like the default, they can change it before the new Fiscal Year begins17:44
phellercedk: sqlite patch works fine17:45
phellertime for a very late breakfast... back in a while17:45
cedkpheller: did you test migration?17:46
cedkudono: ping17:49
-!- chrue( has joined #tryton18:12
cedkudono: did you finish the demo data script?18:19
cedkI have uploaded new 1.6 series for tryton18:22
cedkpheller: here is a command to get last release from hg:18:33
cedkhg up -r `hg tags|head -2|tail -1|cut -d " " -f 1`18:34
-!- FWiesing( has joined #tryton19:34
-!- zodman(~zodman@foresight/developer/zodman) has joined #tryton19:48
phellercedk: shit, I dropped the old database (the one created with neso before the patch).  I'll have to create a new one with trunk, then apply the patch again.19:57
phellercedk: hmmh, it doesn't seem that the sale I created on neso 1.8 before sqlite fix works after I re-run neso with the fix....20:28
cedkpheller: that's possible but is new one works?20:30
phellercedk: if I create a new database from scratch, then yes20:30
cedkpheller: no new sale order20:30
phellercedk: oh oh, let me try it20:30
phellercedk: hmmh, no.  but, the server and module versions are 1.8.0 before the fix, and after the fix... so will the migration be run?20:31
cedkpheller: yes you must run -u all20:32
phellercedk: this will work with neso?20:33
cedkpheller: make a backup and restore it with update checkbox checked20:34
phellercedk: ah ok, trying20:34
-!- udono( has joined #tryton20:36
phellercedk: interesting, the restore database dialog doesn't allow db name to start with digits20:36
cedkpheller: I think there is already an issue for that20:37
phellercedk: hmmh, nope, doesn't seem to work after I reimport the database with update20:37
cedkpheller: what?20:38
phellercedk: I still get the recursion error, even with new sale20:38
phellercedk: let me double check the trytond does have the patch....20:38
phellercedk: d'oh, patch isn't applied.  fixiong20:39
phellercedk: works.20:42
phellercedk: sorry for the confusion20:42
cedkpheller: also for old sale order?20:44
cedkpheller: great20:44
cedkpheller: I will upload soon dmg for old series20:45
phellercedk: ah, you're backporting this patch?20:46
cedkpheller: no20:55
phellercedk: ok... I'm not sure I understand why you need to upload the dmg then?20:56
cedkpheller: to test if they work21:06
-!- chrue( has joined #tryton21:21
-!- enlightx( has joined #tryton22:11
-!- gremly(~gremly@ has joined #tryton22:47
-!- bechamel( has joined #tryton22:54

Generated by 2.11.0 by Marius Gedminas - find it at!