| *** vint_bra has joined #openstack-swift | 00:35 | |
| *** Renich__ has joined #openstack-swift | 00:44 | |
| *** Renich__ is now known as Renich | 00:45 | |
| *** jamielennox is now known as jamielennox|away | 01:01 | |
| *** jamielennox|away is now known as jamielennox | 01:19 | |
| kota_ | good morning | 01:50 |
|---|---|---|
| kota_ | onovy: so cute! congrats! | 01:51 |
| *** vint_bra has quit IRC | 01:52 | |
| mattoliverau | kota_: morning | 02:35 |
| mattoliverau | onovy: \o/ | 02:36 |
| kota_ | mattoliverau: o/ | 02:36 |
| mahatic | good morning | 02:57 |
| mahatic | onovy: congratulations! | 02:57 |
| *** links has joined #openstack-swift | 03:09 | |
| *** sams-gleb has joined #openstack-swift | 03:20 | |
| *** sams-gleb has quit IRC | 03:25 | |
| *** ChanServ sets mode: +v mahatic | 03:28 | |
| *** chosafine has joined #openstack-swift | 03:32 | |
| mattoliverau | mahatic: morning | 03:34 |
| mahatic | mattoliverau: o/ | 03:34 |
| *** chosafine has quit IRC | 03:36 | |
| *** chosafine has joined #openstack-swift | 03:36 | |
| *** psachin has joined #openstack-swift | 04:04 | |
| *** jamielennox is now known as jamielennox|away | 04:14 | |
| *** gkadam has joined #openstack-swift | 04:15 | |
| *** jamielennox|away is now known as jamielennox | 04:22 | |
| *** sams-gleb has joined #openstack-swift | 05:22 | |
| *** rcernin has joined #openstack-swift | 05:26 | |
| *** sams-gleb has quit IRC | 05:27 | |
| *** tone_zrt has joined #openstack-swift | 05:30 | |
| *** ChubYann has quit IRC | 05:32 | |
| *** chosafine has quit IRC | 05:44 | |
| *** oshritf_ has joined #openstack-swift | 05:47 | |
| *** oshritf_ has quit IRC | 05:59 | |
| *** oshritf_ has joined #openstack-swift | 06:10 | |
| *** jaosorior has joined #openstack-swift | 06:13 | |
| *** psachin has quit IRC | 06:33 | |
| *** oshritf_ has quit IRC | 06:38 | |
| *** bkopilov_ has joined #openstack-swift | 06:39 | |
| *** oshritf_ has joined #openstack-swift | 06:45 | |
| *** oshritf_ has quit IRC | 06:50 | |
| *** oshritf_ has joined #openstack-swift | 06:54 | |
| *** jaosorior has quit IRC | 06:58 | |
| *** oshritf_ has quit IRC | 07:03 | |
| *** sams-gleb has joined #openstack-swift | 07:12 | |
| *** sams-gleb has quit IRC | 07:16 | |
| *** jaosorior has joined #openstack-swift | 07:18 | |
| *** pcaruana has joined #openstack-swift | 07:20 | |
| *** tanee is now known as tanee_away | 07:24 | |
| *** tanee_away is now known as tanee | 07:25 | |
| *** tesseract has joined #openstack-swift | 07:44 | |
| *** amoralej|off is now known as amoralej | 07:55 | |
| *** ujjain has quit IRC | 08:00 | |
| *** cbartz has joined #openstack-swift | 08:11 | |
| *** psachin has joined #openstack-swift | 08:12 | |
| *** ujjain has joined #openstack-swift | 08:12 | |
| *** ujjain has joined #openstack-swift | 08:12 | |
| acoles | good morning | 08:20 |
| acoles | onovy: congratulations! enjoy | 08:21 |
| *** PavelK has joined #openstack-swift | 08:26 | |
| *** openstackgerrit has quit IRC | 08:33 | |
| mahatic | acoles: good morning | 08:41 |
| acoles | mahatic: o/ | 08:41 |
| *** gabor_antal_ has quit IRC | 08:51 | |
| *** gabor_antal_ has joined #openstack-swift | 08:52 | |
| *** oshritf has joined #openstack-swift | 09:44 | |
| *** geaaru has joined #openstack-swift | 10:08 | |
| acoles | hi kota_ | 10:12 |
| acoles | mahatic: thanks for looking at patch 449310, that's great. it really is WIP though so expect some changes | 10:18 |
| patchbot | https://review.openstack.org/#/c/449310/ - swift - WIP use composite ring metadata to prevent bad things | 10:18 |
| acoles | mahatic: I'm not yet convinced that the UUID idea is going to be the answer we need | 10:19 |
| mahatic | acoles: sure, np. I would like to test it over, will do that | 10:22 |
| *** jordanP has joined #openstack-swift | 10:22 | |
| jordanP | hi | 10:23 |
| jordanP | according to http://logs.openstack.org/70/449270/1/check/gate-swift-dsvm-functional-ubuntu-xenial/1cdabc2/console.html | 10:23 |
| jordanP | the tox -e func command is ran twice for no reason | 10:23 |
| mahatic | acoles: oic, but md5 definitely seems not to be an option. You could think of conflicts with UUID as well? Per comments from kota_ he seems to be expecting UUID patch | 10:23 |
| jordanP | Ran: 484 tests in 188.0000 sec. and Ran: 484 tests in 176.0000 sec. | 10:23 |
| jordanP | that's because of https://github.com/openstack-infra/project-config/blob/11eb3a36419411b56230ed843e95bb51a4dfcffc/jenkins/jobs/swift.yaml#L33 | 10:24 |
| acoles | jordanP: that's correct | 10:25 |
| jordanP | but it seems weird that the exact same number of tests are ran | 10:25 |
| mahatic | hmm come to think of it, UUID may also face similar issues as md5? I gotta look over some prior comments | 10:25 |
| acoles | jordanP: the two runs use different configs (via the different SWIFT_TEST_CONFIG_FILE you linked to in the yaml) | 10:26 |
| acoles | jordanP: you'll see different number of skipped tests for each run | 10:27 |
| jordanP | acoles, yeah you are right, thanks | 10:28 |
| jordanP | I thooght of a mis config, but all is well | 10:28 |
| *** silor has joined #openstack-swift | 10:28 | |
| jordanP | so I am working on not running every swift-daemon the default gate jobs. Currently devstack does "swift-init all start" and I'd like to selectively enable only PACO + container-sync | 10:29 |
| *** mvk has quit IRC | 10:29 | |
| jordanP | there's no job (no tests: nor functional tests nor tempest tests) that require the entire swift daemon collection | 10:30 |
| jordanP | so I thought maybe we could save some RAM if we just don't run those services | 10:30 |
| acoles | mahatic: potential problem with UUID is (clayg pointed this out to me...) that swift-ring-builder provides a command write_builder to "recover" a builder file from ring file - so we can't assume that a builder file (with UUID burnt into it) will never be lost :/ | 10:31 |
| acoles | mahatic: plus, someone could screw up by copying a builder (with it's UUID) to use as a "template" for another (copy builder file, remove all devs, add new devs) | 10:32 |
| acoles | mahatic: so I'm still pondering it | 10:32 |
| acoles | jordanP: yeah that job definition can be confusing when you first look at it | 10:32 |
| acoles | jordanP: but there is a reason behind it :) | 10:33 |
| jordanP | yep, it's good, makes sense and it's clever | 10:33 |
| acoles | jordanP: the two runs are for keystone auth and v1/tempauth | 10:34 |
| mahatic | acoles: ohh , this "so we can't assume that a builder file will never be lost" seems like a potential blocker. thanks for the info - food for thought :) | 10:37 |
| mahatic | getting my feet wet into the ring land ;) | 10:39 |
| *** silor1 has joined #openstack-swift | 10:40 | |
| *** silor has quit IRC | 10:40 | |
| *** silor1 is now known as silor | 10:40 | |
| *** oshritf has quit IRC | 10:59 | |
| *** zhurong has joined #openstack-swift | 11:31 | |
| *** kei_yama has quit IRC | 11:32 | |
| *** mvk has joined #openstack-swift | 11:44 | |
| *** sams-gleb has joined #openstack-swift | 11:55 | |
| *** oshritf has joined #openstack-swift | 11:58 | |
| *** catintheroof has joined #openstack-swift | 12:08 | |
| *** zhurong has quit IRC | 12:12 | |
| *** tone_z has joined #openstack-swift | 12:15 | |
| *** oshritf has quit IRC | 12:16 | |
| *** tone_zrt has quit IRC | 12:16 | |
| *** amoralej is now known as amoralej|lunch | 12:16 | |
| *** oshritf has joined #openstack-swift | 12:19 | |
| *** gkadam has quit IRC | 12:25 | |
| *** links has quit IRC | 12:40 | |
| *** klamath has joined #openstack-swift | 12:44 | |
| *** chlong has joined #openstack-swift | 12:51 | |
| *** amoralej|lunch is now known as amoralej | 13:00 | |
| *** silor1 has joined #openstack-swift | 13:03 | |
| *** silor has quit IRC | 13:06 | |
| *** silor1 is now known as silor | 13:06 | |
| *** openstackgerrit has joined #openstack-swift | 13:19 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: WIP Use uuid to differentiate component rings in composite https://review.openstack.org/449310 | 13:19 |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Add Composite Ring Functionality https://review.openstack.org/441921 | 13:19 |
| *** silor1 has joined #openstack-swift | 13:33 | |
| *** silor has quit IRC | 13:35 | |
| *** silor1 is now known as silor | 13:35 | |
| jrichli | onovy: congratulations! | 13:38 |
| *** vint_bra has joined #openstack-swift | 13:40 | |
| openstackgerrit | Merged openstack/swift master: Factor out a bunch of common testing setup https://review.openstack.org/446185 | 13:48 |
| *** chosafine has joined #openstack-swift | 13:55 | |
| *** silor1 has joined #openstack-swift | 14:00 | |
| *** gkadam has joined #openstack-swift | 14:00 | |
| *** silor has quit IRC | 14:01 | |
| *** silor1 is now known as silor | 14:01 | |
| *** chlong has quit IRC | 14:09 | |
| *** zhurong has joined #openstack-swift | 14:11 | |
| *** gyee has joined #openstack-swift | 14:13 | |
| *** gyee has quit IRC | 14:13 | |
| *** _JZ_ has joined #openstack-swift | 14:29 | |
| *** silor1 has joined #openstack-swift | 14:30 | |
| *** gkadam has quit IRC | 14:32 | |
| *** silor has quit IRC | 14:32 | |
| *** silor1 is now known as silor | 14:32 | |
| openstackgerrit | Alexandre Lécuyer proposed openstack/swift master: Remove unused returned value object_path from yield_hashes() https://review.openstack.org/450269 | 14:33 |
| *** jordanP has quit IRC | 14:36 | |
| *** jordanP has joined #openstack-swift | 14:36 | |
| *** zhurong has quit IRC | 14:41 | |
| *** szaher has quit IRC | 14:45 | |
| *** szaher has joined #openstack-swift | 14:46 | |
| *** furlongm has quit IRC | 14:50 | |
| *** oshritf has quit IRC | 14:54 | |
| *** rcernin has quit IRC | 15:07 | |
| openstackgerrit | Alexandre Lécuyer proposed openstack/swift master: Remove unused returned value object_path from yield_hashes() https://review.openstack.org/450269 | 15:19 |
| *** silor has quit IRC | 15:20 | |
| *** silor1 has joined #openstack-swift | 15:20 | |
| *** silor1 is now known as silor | 15:22 | |
| *** chlong has joined #openstack-swift | 15:25 | |
| *** cbartz has left #openstack-swift | 15:33 | |
| *** chosafine has quit IRC | 15:42 | |
| *** chsc has joined #openstack-swift | 15:47 | |
| *** chsc has joined #openstack-swift | 15:47 | |
| openstackgerrit | Karen Chan proposed openstack/swift master: Store version id if restoring object from archive https://review.openstack.org/437523 | 15:54 |
| jrichli | Sounds interesting. I plan to review this ^^ - but no promises | 16:05 |
| *** SkyRocknRoll has joined #openstack-swift | 16:16 | |
| *** chsc has quit IRC | 16:19 | |
| *** JimCheung has joined #openstack-swift | 16:22 | |
| *** karenc has joined #openstack-swift | 16:29 | |
| *** mvk has quit IRC | 16:29 | |
| *** psachin has quit IRC | 16:32 | |
| *** jordanP has quit IRC | 16:33 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/swift master: Updated from global requirements https://review.openstack.org/88736 | 16:34 |
| *** jordanP has joined #openstack-swift | 16:34 | |
| *** zaitcev has joined #openstack-swift | 16:36 | |
| *** ChanServ sets mode: +v zaitcev | 16:36 | |
| *** sams-gleb has quit IRC | 16:36 | |
| jordanP | notmyname, what do you think of https://review.openstack.org/#/c/450207/ ? | 16:36 |
| *** sams-gleb has joined #openstack-swift | 16:36 | |
| *** rcernin has joined #openstack-swift | 16:37 | |
| *** sams-gleb has quit IRC | 16:40 | |
| *** jaosorior has quit IRC | 16:41 | |
| *** gyee has joined #openstack-swift | 16:54 | |
| *** sams-gleb has joined #openstack-swift | 16:59 | |
| *** tesseract has quit IRC | 16:59 | |
| *** rcernin has quit IRC | 17:08 | |
| *** chsc has joined #openstack-swift | 17:10 | |
| *** chsc has joined #openstack-swift | 17:10 | |
| *** mvk has joined #openstack-swift | 17:12 | |
| *** pcaruana has quit IRC | 17:12 | |
| *** chsc has quit IRC | 17:16 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Version DLOs, just like every other type of object https://review.openstack.org/446142 | 17:17 |
| *** chlong has quit IRC | 17:22 | |
| *** Renich__ has joined #openstack-swift | 17:25 | |
| *** amoralej is now known as amoralej|off | 17:27 | |
| *** Renich has quit IRC | 17:29 | |
| *** Renich__ is now known as Renich | 17:31 | |
| *** Renich has quit IRC | 17:32 | |
| timburke | good morning! i'm gonna keep chugging away at https://review.openstack.org/#/c/423906/ -- trying to get through the meaty changes in container/backend.py and container/sharder.py! | 17:35 |
| timburke | no patchbot? no... we just haven't fixed the regex yet | 17:35 |
| timburke | https://review.openstack.org/#/c/423906/ | 17:35 |
| patchbot | patch 423906 - swift - Add container sharding to Swift containers | 17:35 |
| jrichli | timburke: its ok, container/sharder.py sort of gave it away :-) | 17:36 |
| *** chlong has joined #openstack-swift | 17:37 | |
| *** tonanhngo has joined #openstack-swift | 17:44 | |
| *** tonanhngo_ has joined #openstack-swift | 17:47 | |
| *** tonanhngo has quit IRC | 17:48 | |
| *** tonanhngo has joined #openstack-swift | 17:50 | |
| *** tonanhngo_ has quit IRC | 17:51 | |
| *** ChubYann has joined #openstack-swift | 17:57 | |
| *** jordanP has quit IRC | 18:02 | |
| *** catintheroof has quit IRC | 18:15 | |
| *** chsc has joined #openstack-swift | 18:15 | |
| *** d0ugal has quit IRC | 18:31 | |
| *** Renich has joined #openstack-swift | 18:37 | |
| notmyname | good morning | 18:43 |
| notmyname | (barely) | 18:43 |
| *** d0ugal has joined #openstack-swift | 18:47 | |
| notmyname | https://review.openstack.org/#/c/450207/ is interesting | 18:50 |
| patchbot | patch 450207 - openstack-dev/devstack - Swift: only start the necessary services | 18:50 |
| notmyname | on the one hand, it's probably right. all those services don't need to be run in a 1-replica policy, and tempest and swift func tests aren't testing them anyway | 18:50 |
| notmyname | on the other, how much of a "real" cluster do we want in the gate | 18:51 |
| notmyname | oh, FYI both clayg and timur are out this week | 18:51 |
| *** PavelK has quit IRC | 19:07 | |
| zaitcev | http://zaitcev.livejournal.com/238656.html VMware OpenStack? I'm curious if they have Swift. VMware owns EMC or vice versa, so that's a ready-made storage business. | 19:22 |
| *** samueldmq has quit IRC | 19:23 | |
| *** samueldmq has joined #openstack-swift | 19:24 | |
| *** silor has quit IRC | 19:24 | |
| notmyname | zaitcev: from what I've seen, most big packaged openstack products (like vmware and even red hat to an extent) have had "(most of the rest of) OpenStack! plus our own in-house storage system with a Swift API" | 19:24 |
| openstackgerrit | Nelson Marcos de Almeida proposed openstack/python-swiftclient master: Removing duplicated doc from client-api https://review.openstack.org/450426 | 19:25 |
| *** NM has joined #openstack-swift | 19:26 | |
| *** jordanP has joined #openstack-swift | 19:29 | |
| notmyname | jordanP: oh, I totally agree that 350MB is something. just seems like it's unlikely to be a large percentage of what's being used. of course, a bunch of small things can really matter, too | 19:34 |
| jordanP | notmyname, we've been chasing memory consumption for a month now, and we kinda lack of fresh ideas. I get your point, and I agree, but 350 Mo of out 8Go, is something. If you have any idea where to save some more Mo, that would be great | 19:35 |
| notmyname | jordanP: to be clear, I'm agreeing with you | 19:36 |
| jordanP | we tried mysql here: https://review.openstack.org/#/c/438668/ but was reverted | 19:36 |
| jordanP | because of some side effects | 19:36 |
| jordanP | we have a nice tool that prints regularly which process is consuming what amount of memory. for instance: http://logs.openstack.org/68/438668/5/check/gate-tempest-dsvm-neutron-full-ubuntu-xenial/a3230f9/logs/screen-peakmem_tracker.txt.gz | 19:38 |
| jordanP | fwiw | 19:38 |
| notmyname | interesting | 19:39 |
| *** zacksh has quit IRC | 19:41 | |
| *** zacksh has joined #openstack-swift | 19:42 | |
| notmyname | if any of us in our products were getting OOM errors and only had 8GB of memory, we'd simply buy more RAM. it's by far cheaper ($ and opportunity cost) than having a team of devs looking for every scrap of memory usage to trim down | 19:43 |
| notmyname | however, when the hardware is donated (a la infra cloud) and the cost is to an open source community, the incentives aren't aligned to solve it with more hardware | 19:43 |
| notmyname | so we end up getting a global community to spend months (cumulatively, if not literally) debugging OOM errors and finding places to save | 19:44 |
| jordanP | yeah, honestly, we've been running with 8Gb of ram since 2015, so maybe we could ask for an increase now | 19:44 |
| jordanP | but we have more and more projects, and more and more jobs to run | 19:44 |
| notmyname | yep | 19:44 |
| notmyname | "run every cloud infrastucture service in a VM simulating a cloud test environment, and do it in 8GB of RAM" seems a little aggressive to me :-) | 19:45 |
| jordanP | yeah, I think the same. But I don"t know what else to do. | 19:46 |
| notmyname | containers! | 19:46 |
| notmyname | (of course) | 19:46 |
| jordanP | many projects suffer from a high rate of false negative | 19:46 |
| jordanP | it's really hurting | 19:46 |
| notmyname | I'm only half joking about the containers thing | 19:47 |
| notmyname | mostly about the "containers" part of it | 19:47 |
| jordanP | I am not sure I see how containers would help here. every container will run on the same host | 19:48 |
| notmyname | but running different services on different nodes seems to be much closer to reality (and significantly raises the amount of HW available per service) | 19:48 |
| jordanP | devstack is not super equiped to do multinode testing yet, and it will means more test VMs | 19:54 |
| *** SkyRocknRoll has quit IRC | 20:04 | |
| *** jamielennox is now known as jamielennox|away | 20:13 | |
| *** catintheroof has joined #openstack-swift | 20:23 | |
| *** jamielennox|away is now known as jamielennox | 20:24 | |
| *** catintheroof has quit IRC | 20:39 | |
| clarkb | I mean we do have multinode testing | 20:47 |
| clarkb | (and I'd like to delete our single node base jobs and replace with multinode but I think that scares some people due to complexity) | 20:48 |
| notmyname | clarkb: ironic that openstack devs would be scared of texting complexity for openstack ;-) | 20:49 |
| clarkb | in their defense a lot of the complexity and "external" to openstack in that you have to set up networking and things in clouds that don't let you control that directly | 20:53 |
| clarkb | but definitely things like live migration etc have had openstack specific complexity issues there (it turns out that we were enforcing cpu flags on qemu booted VMs for a long time when we didn't need to because its qemu and not kvm) | 20:54 |
| *** jordanP has quit IRC | 21:22 | |
| *** lcurtis has joined #openstack-swift | 21:25 | |
| lcurtis | hello everyone...looking at changing standard drive size in our swift object servers from 4T to 10T...concerned that these higher-weighted drives will get 2x the I/O. Is anyone currently running with mixed drive sizes? | 21:26 |
| *** geaaru has quit IRC | 21:29 | |
| mattoliverau | Morning | 21:31 |
| mattoliverau | timburke: thanks for looking at sharding btw! It's great to have your eye over it cause I find myself somewhat shard blind to the code now :p | 21:41 |
| timburke | mattoliverau: yeah, happy to! we all want it, most of us need it, and you've been working on it a while. let's get it landed! | 21:41 |
| notmyname | mattoliverau: yep I'm just now talking to orion in the office about it :-) | 21:43 |
| mattoliverau | \o/ | 21:43 |
| *** NM has quit IRC | 21:44 | |
| mattoliverau | I know there is some ways to go, but glad its finally not a POC anymore.. but that also means there are defintley places we can clean up or do better. I put a question on the trello regarding how to send new pivot ranges (after sharding).. that might bypass .pending for changes to the pivot ranges which could be good. Just FYI | 21:45 |
| *** vint_bra has quit IRC | 21:49 | |
| *** sams-gleb has quit IRC | 22:02 | |
| *** chlong has quit IRC | 22:03 | |
| notmyname | lcurtis: hey. sorry for the delay | 22:15 |
| notmyname | lcurtis: you've just explained why I'm so scared about all these bigger drives hitting the market. they're a *lot* bigger, but the bus isn't any faster and they have essentially the same IOPS. terrifying! | 22:15 |
| lcurtis | notmyname: agreed! | 22:16 |
| notmyname | lcurtis: most clusters will end up with mixed drive sizes over time. basically, you end up adding whatever drive size is the best $/GB at the moment | 22:16 |
| notmyname | lcurtis: but I've always seen one drive size per box (but different in different boxes) | 22:17 |
| lcurtis | do u think it a problem to have disparate drive sizes? it would seem to generate hotspots in the cluster | 22:17 |
| lcurtis | ie, a rack of servers with 10T drives , higher weight | 22:17 |
| lcurtis | more I/O on those servers? | 22:17 |
| notmyname | no, I don't think it's a problem in general. larger drives (8-10TB) make dense nodes more scary to me (those 60, 80 bay chassis) | 22:18 |
| notmyname | regardless of drive sizes, there are other things we're working on in swift to fix hot spot issues :-) | 22:18 |
| lcurtis | awesome | 22:18 |
| lcurtis | pure magic | 22:18 |
| notmyname | that's the impetus behind a lot of the replication/reconstruction and golang work you may have heard about | 22:18 |
| notmyname | but it's a lot of work, and a long-term project | 22:19 |
| notmyname | but back to your current question... big drives | 22:19 |
| notmyname | I personally like smaller nodes as drives get bigger. but then clayg gripes at me for solving problems with hardware and I feel bad. (but I still like less dense nodes myself) | 22:20 |
| notmyname | lcurtis: it depends on your workload (lots of small files vs lots of big files vs high reads vs high writes vs etc), but the biggest problems you'll see with big drives is with background processes, not client performance | 22:21 |
| notmyname | lcurtis: but the specific answer to your question is that a cluster with mixed drive sizes is totally normal, expected, and supported | 22:22 |
| lcurtis | like auditor, etc? | 22:22 |
| notmyname | auditing, replication, reconstruction, etc | 22:22 |
| lcurtis | makes sense | 22:22 |
| lcurtis | but wouldnt a device with higher weight see more i/o than peers with lower weight? | 22:23 |
| notmyname | yes | 22:23 |
| lcurtis | we are pretty maxed with i/o at the moment on object tier | 22:23 |
| notmyname | add more ;-) | 22:23 |
| lcurtis | okay...i just wanted to make sure my thinking was in line | 22:23 |
| notmyname | to clarify (especially for the searchable channel logs) the "issues" or "problems" here are relative. swift is still awesome at big clusters, lots of data, and lots of traffic | 22:23 |
| lcurtis | yes..thinking was to add more 10t nodes, backfill rest of cluster | 22:24 |
| lcurtis | of 4ts | 22:24 |
| lcurtis | but that seems like time consuming endeavor | 22:24 |
| notmyname | and although the drives are 5x the size of when the project started (2TB vs 10TB), that's a relatively small multiple. we're not talking orders of magnitude differences like you'd get with flash | 22:24 |
| lcurtis | true | 22:24 |
| notmyname | where is the disk IO going today? is it in the client requests or in stuff like rsync or the background daemons? | 22:25 |
| lcurtis | just sheer puts | 22:25 |
| lcurtis | and rebalancing | 22:26 |
| *** gkadam has joined #openstack-swift | 22:26 | |
| notmyname | cool (and congrats) :-) | 22:26 |
| lcurtis | at the moment | 22:26 |
| notmyname | yeah, the rebalancing is the major io issue that's being addressed for dense nodes | 22:26 |
| notmyname | rebalancing = either EC reconstruction or replication or both (just "rebalancing" is a lot shorter to type) | 22:27 |
| notmyname | the good news for the PUT load is that it should go up pretty smoothly with more spindles | 22:27 |
| notmyname | you'll need more tuning for getting rebalancing to go faster | 22:28 |
| lcurtis | good stuff as always, notmyname | 22:34 |
| lcurtis | thank you | 22:34 |
| *** sams-gleb has joined #openstack-swift | 23:02 | |
| *** sams-gleb has quit IRC | 23:08 | |
| openstackgerrit | Merged openstack/python-swiftclient master: Removing duplicated doc from client-api https://review.openstack.org/450426 | 23:21 |
| *** lcurtis has quit IRC | 23:24 | |
| *** kei_yama has joined #openstack-swift | 23:29 | |
| notmyname | I have added all of the session topics mentioned on https://etherpad.openstack.org/p/BOS-Swift-brainstorming to the formal submission session | 23:45 |
| notmyname | (I think) you can see all the submitted stuff at http://forumtopics.openstack.org | 23:45 |
| notmyname | at least I can | 23:45 |
| notmyname | there's an opportunity to comment on any of the suggested sessions. please don't hesitate to add comments | 23:46 |
| notmyname | I suspect we' | 23:46 |
| *** chsc has quit IRC | 23:46 | |
| notmyname | I suspect we'll need to advocate for and show community support for the swift sessions | 23:46 |
| notmyname | and if there's another session you want to see submitted, please let me know. | 23:47 |
| notmyname | I love swob https://review.openstack.org/#/c/423366/ | 23:49 |
| notmyname | https://review.openstack.org/#/c/423366/ | 23:50 |
| patchbot | patch 423366 - glance - Fix incompatibilities with WebOb 1.7 | 23:50 |
| *** klamath has quit IRC | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!