*** vinsh has quit IRC | 00:03 | |
*** dmsimard is now known as dmsimard_away | 00:05 | |
*** rdaly2 has quit IRC | 00:18 | |
*** zhill has joined #openstack-swift | 00:23 | |
*** os1 has quit IRC | 00:28 | |
zaitcev | clayg: There was a patch from one of our Japanese crew guys just recently about this. The old behavior was, if container updater gets 404, just drop the update. His idea was to retry forever. I do not remember what the justification was. Apparently some kind of inconsistency. | 00:35 |
---|---|---|
zaitcev | clayg: I don't remember if it went it. I wasn't a fan. | 00:35 |
zaitcev | Actually.... | 00:35 |
zaitcev | https://review.openstack.org/99598 - here it is. It's the younger guy, Takahashi. "for example when all containers are putted at handoff devices", he says. | 00:37 |
*** dmorita has joined #openstack-swift | 00:38 | |
*** lcurtis has quit IRC | 00:41 | |
clayg | zaitcev: I'm not sure it's specifically related, I just wanted to retry till reclaim_age - redbo told me recently that he'd seen really old async pendings sneak an object update back into a container long after the object had been tombstoned and the deleted row in the container reaped | 00:43 |
zaitcev | clayg: well, that makes sense | 00:44 |
clayg | zaitcev: it's sort of a wonky situation, because normally the tombstone would have async'd over the put - the object updater does something to compress those async records - but if the part was also handed off the tombstone record could have gotten processed from another node | 00:44 |
clayg | then it's a question why this one node's object-updater wasn't doing his asyncs | 00:45 |
zaitcev | okay, that's a level down. imkarrer is dealing with container updater | 00:45 |
clayg | zaitcev: yeah I think so | 00:45 |
*** dmsimard_away is now known as dmsimard | 00:52 | |
*** david-lyle is now known as david-lyle_afk | 00:59 | |
*** nellysmitt has joined #openstack-swift | 01:03 | |
*** nellysmitt has quit IRC | 01:07 | |
*** zhill has quit IRC | 01:09 | |
*** zhill_ has joined #openstack-swift | 01:10 | |
*** abhirc has quit IRC | 01:17 | |
*** chlong has quit IRC | 01:40 | |
*** chlong has joined #openstack-swift | 01:41 | |
*** kajinamit has joined #openstack-swift | 01:45 | |
*** abhirc has joined #openstack-swift | 01:53 | |
*** vinsh has joined #openstack-swift | 02:01 | |
*** vinsh has quit IRC | 02:21 | |
*** dmsimard is now known as dmsimard_away | 02:23 | |
*** jrichli has joined #openstack-swift | 02:29 | |
*** NM has joined #openstack-swift | 02:32 | |
clayg | oh wait, so we didn't come to any concensus on publishing the reseller prefix, except that glange and I both think it's weird - and notmyname is out of town till next week. | 02:51 |
mattoliverau | I haven't properly reviewed that change yet, still trying to install keystone :( | 02:55 |
clayg | mattoliverau: apparently acoles has a script for that? he said something about using his saio with a stand alone keystone endpoint | 02:55 |
clayg | which I like very much except the part where I have to maintain a keystone provisioning script | 02:56 |
clayg | but if *acoles* maintained it!?! | 02:56 |
mattoliverau | I'll go back and look at his script.. at first glance it configured keystone not set it up.. but I could be wrong, besides if it does that is still decreases the effort required :) | 02:57 |
clayg | or mattoliverau - I'm not picky - basically anything in the the difference of the set of all people in the world and set([me]) | 02:57 |
mattoliverau | clayg: lol, i like the way you think.. thanks for volunteering to keep it maintained acoles_away :P | 02:57 |
clayg | well that gist he linked was just about provisioning users against a devstack keystone endpoint since it doesn't come setup will all the service roles and the like by default | 02:58 |
clayg | I think most of my problems with devstack may truely have been just that I *only* provisioned 1GB of memory for the poor thing | 02:59 |
mattoliverau | Sigh.. I don't have the required launchpad rights to do triage or invalidate bugs, I guess I need to go complain to infra | 02:59 |
clayg | which seemed like enough since I was only running swift and keystone and I've successfully run saio's on 512 in the past - I had assumed that a keystone deployment would take less memory than swift (oh i'm so silly!) | 02:59 |
mattoliverau | clayg: lucky I have a rackspace dev account ;) | 03:00 |
mattoliverau | lol | 03:00 |
clayg | oh yeah, notmyname also swears by saio's in the cloud - I've done that before I suppose | 03:00 |
mattoliverau | I maybe one of the biggest Sydney region users.. and I don't pay. But its what I do, I keep the syd dcops guys employed :P | 03:02 |
*** nellysmitt has joined #openstack-swift | 03:03 | |
*** nellysmitt has quit IRC | 03:08 | |
*** NM has quit IRC | 03:14 | |
*** panbalag has quit IRC | 03:18 | |
*** vinsh has joined #openstack-swift | 03:22 | |
*** zhill_ has quit IRC | 03:22 | |
*** vinsh has quit IRC | 03:26 | |
*** gyee has quit IRC | 03:28 | |
*** vinsh has joined #openstack-swift | 03:53 | |
*** NM has joined #openstack-swift | 04:20 | |
*** zaitcev has quit IRC | 04:36 | |
*** SkyRocknRoll has joined #openstack-swift | 04:40 | |
*** lcurtis has joined #openstack-swift | 05:04 | |
*** nellysmitt has joined #openstack-swift | 05:05 | |
*** ppai has joined #openstack-swift | 05:05 | |
lcurtis | hello all...for some reason i was reading an article that warned against using built in python web server in swift for larger environments...instead using something else like haproxy or apache...am i mistaken? | 05:06 |
*** nellysmitt has quit IRC | 05:09 | |
*** jrichli has quit IRC | 05:11 | |
clayg | lcurtis: all of the largest deployments I know about use the built in python web server - the thing is you can't terminate ssl in it | 05:22 |
clayg | so stud or haproxy for ssl termination with the built in swift/eventlet proxy processes listening on your localhost is generally gunna treat ya | 05:23 |
clayg | for the backend services, you can just bind stright up to the private/cluster network interfaces | 05:23 |
clayg | i guess I've heard of people doing something with apache, and one time this one guy tried something with nginx but it didn't work out | 05:23 |
*** SkyRocknRoll has quit IRC | 05:26 | |
clayg | ahale: we switched to stud for ssl termination cause like one time syn said it was great, and generally speaking it's been better than haproxy, but stupid poodle was a disaster because of https://github.com/bumptech/stud/pull/138/files not being merged | 05:26 |
clayg | ahale: we were gunna fork and repackage (as we do) but swifterdarrell found this crazy shit -> https://github.com/Netfuture/tlsinterposer | 05:28 |
clayg | so you LD_PRELOAD this shim that lets you get the right f'ing SSL_CTX options you want | 05:28 |
clayg | I breifly argued it would be easier to just patch stud like the rest of world, but swifterdarrell had convinced himself it would be a good thing to have around the next time we have some stupid glue app that won't let us setup SSL the way we want | 05:29 |
clayg | i always want to call it the tls-interloper | 05:32 |
*** SkyRocknRoll has joined #openstack-swift | 05:42 | |
*** NM has quit IRC | 05:42 | |
clayg | that moment when you realize your "quick fix" has become an 800 line diff | 05:44 |
*** lcurtis has quit IRC | 05:44 | |
clayg | i'll see what i can do to chop it up later | 05:44 |
clayg | g'night | 05:44 |
*** mrda is now known as mrda-weekend | 06:01 | |
*** ahonda has quit IRC | 06:16 | |
*** kajinamit has quit IRC | 06:26 | |
*** nshaikh has joined #openstack-swift | 06:44 | |
*** nellysmitt has joined #openstack-swift | 07:05 | |
*** nellysmitt has quit IRC | 07:10 | |
*** SkyRocknRoll has quit IRC | 07:14 | |
*** silor has joined #openstack-swift | 07:19 | |
*** jamielennox is now known as jamielennox|away | 07:22 | |
*** nellysmitt has joined #openstack-swift | 07:23 | |
*** SkyRocknRoll has joined #openstack-swift | 07:31 | |
*** nshaikh has left #openstack-swift | 07:53 | |
ahale | oO | 07:56 |
*** nshaikh has joined #openstack-swift | 08:02 | |
*** ahonda has joined #openstack-swift | 08:08 | |
*** mmcardle has joined #openstack-swift | 08:09 | |
*** rledisez has joined #openstack-swift | 08:12 | |
*** madhuri_ has quit IRC | 08:33 | |
*** jistr has joined #openstack-swift | 08:46 | |
*** madhuri has joined #openstack-swift | 08:47 | |
*** ppai has quit IRC | 09:03 | |
*** madhuri has quit IRC | 09:10 | |
*** ppai has joined #openstack-swift | 09:15 | |
*** jordanP has joined #openstack-swift | 09:19 | |
*** khivin has joined #openstack-swift | 09:26 | |
*** ppai has quit IRC | 09:28 | |
*** ppai has joined #openstack-swift | 09:40 | |
*** acoles_away is now known as acoles | 09:54 | |
acoles | clayg: mattoliverau : so it did cross my mind to script everything from bare VM to swift-config'd-keystone. i could do it while waiting for gerrit to serve pages ;) | 10:08 |
*** aix has joined #openstack-swift | 10:12 | |
acoles | clayg: i run keystone in a 1GB vm, devstack in 4GB (lucky choice i guess). what did they use to land on the moon? :o | 10:18 |
openstackgerrit | Alistair Coles proposed openstack/swift-specs: Updates to encryption spec https://review.openstack.org/154318 | 10:28 |
*** MasterPiece has joined #openstack-swift | 10:30 | |
*** SkyRocknRoll has quit IRC | 10:48 | |
openstackgerrit | Merged openstack/swift: Update guest VM OS recommendation in SAIO doc https://review.openstack.org/156539 | 10:56 |
*** ppai has quit IRC | 11:11 | |
*** ppai has joined #openstack-swift | 11:25 | |
*** jasondotstar has joined #openstack-swift | 11:32 | |
*** openstack has joined #openstack-swift | 20:23 | |
*** zhill has quit IRC | 20:24 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!