Friday, 2019-03-01

*** itlinux_ has quit IRC00:13
*** mvkr has joined #openstack-swift00:14
*** itlinux has joined #openstack-swift00:18
timburkezaitcev, fyi: we can move forward on ssync as long as we've got new enough eventlet: https://github.com/eventlet/eventlet/commit/4b0450b37c26d77cebf58d15f6c07525b5073b7200:35
timburkealso, i might be getting close to being able to run some func tests since eventlet landed https://github.com/eventlet/eventlet/commit/f0bc79e1d973806866e2eb74db82adfda64ab1e5 -- gotta use latin1 around https://github.com/python/cpython/blob/v3.7.2/Lib/http/client.py#L1106-L1107 though00:36
*** itlinux has quit IRC00:44
*** itlinux has joined #openstack-swift01:09
*** itlinux has quit IRC01:11
*** itlinux has joined #openstack-swift01:12
*** itlinux has quit IRC01:18
*** itlinux has joined #openstack-swift01:19
*** henriqueof has quit IRC01:30
*** itlinux has quit IRC01:51
*** itlinux has joined #openstack-swift01:52
*** itlinux has quit IRC01:58
*** tkajinam has quit IRC02:02
*** tkajinam has joined #openstack-swift02:03
*** tkajinam_ has joined #openstack-swift02:05
*** tkajinam has quit IRC02:07
*** tkajinam_ has quit IRC02:10
*** tkajinam has joined #openstack-swift02:10
*** gyee has quit IRC02:42
*** psachin has joined #openstack-swift03:12
openstackgerritMerged openstack/swift master: py3: port object auditor  https://review.openstack.org/63919803:51
zaitcevOh goodness gracious, at last05:03
*** tingjie has joined #openstack-swift05:40
*** tingjie has left #openstack-swift05:40
*** tingjie has joined #openstack-swift05:41
*** tingjie has left #openstack-swift05:41
*** itlinux has joined #openstack-swift06:33
*** itlinux_ has joined #openstack-swift06:37
*** itlinux has quit IRC06:40
*** zaitcev has quit IRC07:23
*** e0ne has joined #openstack-swift07:28
*** itlinux_ has quit IRC07:37
*** gkadam has joined #openstack-swift07:37
*** itlinux has joined #openstack-swift07:39
*** itlinux has quit IRC07:55
*** tkajinam has quit IRC08:18
*** ccamacho has joined #openstack-swift08:19
*** mcape has joined #openstack-swift09:35
mcapeHi guys! I just read swift upgrade notes from the Ocata - 2.11.0,  till Rocky - 2.19.0.  I have impression that I can upgrade code of swift directly from Ocata to Rocky passing intermediate versions.09:38
mcapeCan anyone confirm that? What are your feelings? :-)09:38
mcapeThere is a 20+ mln file container in that Ocata deployment, and it is under heavy HEAD load10:01
mcapeso, there is a need to figure out how to execute rapid transition to version which supports container sharding10:09
*** e0ne has quit IRC10:22
*** e0ne has joined #openstack-swift10:23
*** mcape has quit IRC10:29
*** kazsh has quit IRC10:46
*** kazsh has joined #openstack-swift10:52
*** e0ne has quit IRC11:11
*** e0ne has joined #openstack-swift11:14
*** mcape has joined #openstack-swift11:19
*** e0ne has quit IRC11:49
*** henriqueof has joined #openstack-swift11:55
*** mcape has quit IRC12:05
*** mcape has joined #openstack-swift12:10
*** threestrands has quit IRC12:24
*** ybunker has joined #openstack-swift12:39
ybunkerhi to all, quick question... i just deploy two new data nodes to an already created cluster, and in the rsyncd.log file im getting the following errors messages:12:40
ybunkerunknown module 'container' tried from (hostname) xx.xx.xx.xx12:40
ybunkerand i notice that objects are written to the disks.. but accounts or containers are not12:41
ybunkeralso on the account logs i find this:12:48
ybunkeraccount-replicator: Can't find itself 127.0.0.1, ::1, 172.xx.xx.xx, xxxx::xxxx:xxxx:xxxx:xxxx, 10.xx.xx.xx, xxxx::xxxx:xxxx:xxxx:xxxx, 172.xx.xx.xx, xxxx::xxxx:xxxx:xxxx:xxxx with port 4101 in ring file, not replicating12:50
*** e0ne has joined #openstack-swift12:50
*** ybunker has quit IRC13:03
*** ybunker has joined #openstack-swift13:04
ybunkerany ideas?13:06
*** ybunker has quit IRC13:21
*** ybunker has joined #openstack-swift13:25
ybunkeranyone?13:27
*** ybunker has quit IRC13:33
*** ccamacho has quit IRC14:33
*** ccamacho has joined #openstack-swift14:34
*** ccamacho has quit IRC14:51
*** ccamacho has joined #openstack-swift14:53
*** itlinux has joined #openstack-swift15:01
*** itlinux has quit IRC15:12
*** mcape has quit IRC15:18
*** itlinux has joined #openstack-swift15:29
*** itlinux has quit IRC15:36
*** itlinux has joined #openstack-swift15:38
*** jistr is now known as jistr|mtg15:38
*** e0ne has quit IRC15:38
*** e0ne has joined #openstack-swift15:39
notmynamegood morning15:50
*** jistr|mtg is now known as jistr16:02
*** e0ne has quit IRC16:14
*** e0ne has joined #openstack-swift16:15
*** e0ne has quit IRC16:17
*** gkadam has quit IRC16:27
openstackgerritJohn Dickinson proposed openstack/python-swiftclient master: authors/changelog updates for release  https://review.openstack.org/63947416:33
notmynametimburke: should be a little more correct and clear ^^16:33
notmynametimburke: and I've gone ahead and added a +2/+A to get the gate started. yell at me if something is wrong :-)16:34
*** psachin has quit IRC16:38
*** itlinux has quit IRC16:49
*** gyee has joined #openstack-swift17:04
timburkenotmyname, worth mentioning p 639777?17:39
patchbothttps://review.openstack.org/#/c/639777/ - python-swiftclient - Add py37 check/gate jobs; add py37 to default tox ... (MERGED) - 2 patch sets17:39
*** mvkr has quit IRC18:02
*** mvkr has joined #openstack-swift18:06
*** irclogbot_2 has joined #openstack-swift18:12
notmynametimburke: IMO, not really. if there had been some change we had to make to the code to support py3 (ie it definitely didn't work before), then yes absolutely we'd mention it. but since it's a "now we're actually testing that" but didnt' change anything, it's not as big of a deal to call out18:23
*** Chealion has joined #openstack-swift18:43
*** e0ne has joined #openstack-swift19:12
*** e0ne has quit IRC19:31
*** irclogbot_2 has quit IRC19:49
*** irclogbot_2 has joined #openstack-swift20:04
*** e0ne has joined #openstack-swift20:05
*** guimaluf has joined #openstack-swift20:37
*** zaitcev has joined #openstack-swift20:39
*** ChanServ sets mode: +v zaitcev20:39
*** chocolate-elvis has joined #openstack-swift20:48
henriqueofnotmyname: Can I deploy swift using a LVM LV?21:09
notmynamehenriqueof: yes, in the sense that it will work. format your LV with one partition, put XFS on it, and you're good to go21:09
notmynamehowever...21:09
notmynameit's generally not a good idea to run in prod that way. or rather, it's probably not a good idea21:10
notmynamefor example, you don't want to have a bunhc of physical volumes in the volume group you're carving logical volumes out of21:10
notmynameit's very important to not hide physical failure domains (drives, servers, racks, DCs) from swift. swift uses that info to make placement decisions to protect durability and availability21:11
notmynameif you start hiding them (eg several logical volumes backed by one physical volume), bad things may happen. for example, swift may place replicas on two separate drives, but if they're backed by the same PV, and it dies, you're in a much worse state than if it just had one replica on it21:12
notmynameso... to set up a dev cluster for some testing? sure, why not? if it makes it easier, go for it. (TBH, I have my dev box set up similarly to this)21:13
notmynamefor prod data? I think that's a bad idea21:13
henriqueofOkay, but I have a layer of integrity, I am doing it on a Dell R730xd, it have a RAID controller so if one disk fails it wont affect the deployment immediately.21:14
henriqueofWhat dou you think?21:15
notmynameI think you're adding a bunch of layers of complexity that will, at best, make debugging hard, and at worst start competing against each other.21:17
notmynamethe best way to deploy swift is on bare metal machines with a bunch of direct-attached drives. no raid, no virtualization, no lvm. just a bunch of disks21:17
henriqueofI see, don't get me wrong, I am dealing with a lack of resources, need to work with what I have. :)21:18
notmynameyeah, I understand :-)21:19
notmynameI'm not saying it won't *work*21:19
henriqueofJust need a object storage solution and I don't think Ceph goes well with only one node.21:19
notmynamehow many drives do you have in that server?21:19
henriqueof17 x 1.2T, 10K21:20
openstackgerritTim Burke proposed openstack/swift master: py3: fix copying unicode names  https://review.openstack.org/64051821:22
openstackgerritTim Burke proposed openstack/swift master: Clean up func tests ahead of py3  https://review.openstack.org/64051921:24
notmynamehenriqueof: are those just data drives? do you have any flash drives in the system?21:26
henriqueofnotmyname: All of then are data drives.21:27
notmynamehenriqueof: adding all 17 drives with a 2x or 3x replica policy (depending on the durability, overhead you need) will work just fine. no need for adding RAID or LVM or anything like that21:31
henriqueofWell, I need to run Nova and Cinder on this server too.21:34
*** irclogbot_2 has quit IRC21:38
henriqueofHow about I remove then from RAID, leave some physical disks for Swift and add the remaining on a big VG?21:39
notmynameah. um... sure? that will probably work21:40
notmynameyou've got a limited set of hardware that you're trying to do a lot of stuff with. so it's all a matter of where you want to make the tradeoffs.21:40
notmynameI mean, everything will function. it's just a question of what you prioritize on the limited hardware21:41
*** e0ne has quit IRC22:24
*** henriqueof has quit IRC22:27
openstackgerritMerged openstack/python-swiftclient master: authors/changelog updates for release  https://review.openstack.org/63947422:29
notmynamehttps://review.openstack.org/64054922:39
patchbotpatch 640549 - releases - swiftclient 3.6.1 release - 1 patch set22:39
*** nottrobin_ has joined #openstack-swift22:45
*** beisner_ has joined #openstack-swift22:45
*** seongsoocho_ has joined #openstack-swift22:45
*** seongsoocho has quit IRC22:52
*** beisner has quit IRC22:52
*** nottrobin has quit IRC22:52
*** beisner_ is now known as beisner22:52
*** seongsoocho_ is now known as seongsoocho22:52
*** nottrobin_ is now known as nottrobin22:52
*** kazsh has quit IRC22:55
*** kazsh has joined #openstack-swift22:59
openstackgerritTim Burke proposed openstack/swift master: Stop monkey-patching mimetools  https://review.openstack.org/64055223:04
*** frickler has quit IRC23:06
*** frickler has joined #openstack-swift23:06
zaitcevtimburke: coincidentally, I had a hand-rolled unquote() in bulk. It was the only way to do it... I'll investigate if I can use your wsgi_unquote in swob. Check this out: https://review.openstack.org/#/c/619303/2/swift/common/utils.py23:08
patchbotpatch 619303 - swift - py3: port bulk middleware - 2 patch sets23:08
notmynametimburke: I'm not sure on p 640552. it doesn't seem as reliable. maybe that's because I don't really trust eventlet to keep a totally stable interface there? I cetainly don't have a better idea, especially given that the py2 mechanism just doesn't exist in py323:10
patchbothttps://review.openstack.org/#/c/640552/ - swift - Stop monkey-patching mimetools - 1 patch set23:10
timburkezaitcev, oh jeez... i think i only got as far as noticing the quote_x()...23:11
timburkenotmyname, there's a thing you could monkey patch to have a similar effect... it's just over in email.message, and when you touch it, pkg_resources breaks23:11
notmynamehow/why does it break?23:12
notmyname /curious23:12
timburkemeanwhile, eventlet hasn't touched https://github.com/eventlet/eventlet/blob/v0.24.1/eventlet/wsgi.py#L646-L652 in ~5years so... yay?23:12
notmynameseems like it's time to touch it!! ;-)23:13
timburkelemme get the traceback... it was extraordinarily not-obvious23:13
timburkenotmyname, http://paste.openstack.org/show/746635/23:13
notmynameyeah, there is nothing in there that says "you changed the email module"23:15
timburkenotmyname, the good news is, i reworked https://review.openstack.org/#/c/640552/1/test/unit/proxy/test_server.py to do an end-to-end test, so we'll know if/when it breaks23:16
patchbotpatch 640552 - swift - Stop monkey-patching mimetools - 1 patch set23:16
timburke(of course, we'd also get that out of our func tests, but y'know...)23:16
notmynameit would almost seem that there some sort of hashing of dependencies?23:17
timburkeit's got soemthing to do with the parsing going on in https://github.com/pypa/setuptools/blob/master/pkg_resources/__init__.py#L2670-L270923:19
timburke(i think)23:19
timburkesee also: https://github.com/pypa/setuptools/blob/master/pkg_resources/__init__.py#L296923:20
timburkepersonally, i rather like the change. more than twice as many deletions as insertions, change is localized to the bit that actually cares about it... needing to remember to pass the SwiftHttpProtocol in test/unit/helpers.py was a little annoying but ultimately a better mirror of reality. the only real badness is the coupling with eventlet23:23
timburkeit's almost like it'd be nice to get off eventlet or something...23:23
notmynameheh23:23
zaitcevWhat's the alternative?23:39
zaitcevThere was something in py3, but I forgot its moniker.23:40
zaitcevMaybe this? http://bottlepy.org/docs/dev/23:41
openstackgerritTim Burke proposed openstack/swift master: py3: port object updater  https://review.openstack.org/63920123:41
zaitcev"Bottle supports Python 2.7 and Python 3."23:41
zaitcevI don't want the cure to be worse than the disease.23:42
notmynamesweventlet, of course ;-)23:43
zaitcevoh go23:53
zaitcevd23:53
zaitcevI think we should have a recheck deadpool. The object auditor took 5 rechecks. So.... Let's give updater 4.23:54

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!