| *** strobert1 has quit IRC | 00:32 | |
| *** strobert1 has joined #openstack-swift | 01:10 | |
| *** nanzha has joined #openstack-swift | 01:24 | |
| *** spsurya has joined #openstack-swift | 01:40 | |
| *** threestrands has quit IRC | 05:51 | |
| *** psachin has joined #openstack-swift | 06:09 | |
| *** pcaruana has joined #openstack-swift | 06:34 | |
| *** nanzha has quit IRC | 06:51 | |
| *** nanzha has joined #openstack-swift | 06:51 | |
| *** nanzha has quit IRC | 07:21 | |
| *** nanzha has joined #openstack-swift | 07:31 | |
| *** nanzha has quit IRC | 08:00 | |
| *** nanzha has joined #openstack-swift | 08:01 | |
| *** tesseract has joined #openstack-swift | 08:04 | |
| *** rdejoux has joined #openstack-swift | 08:13 | |
| *** tkajinam has quit IRC | 08:21 | |
| *** nanzha has quit IRC | 08:41 | |
| *** nanzha has joined #openstack-swift | 08:43 | |
| *** mikecmpbll has joined #openstack-swift | 08:51 | |
| *** rpittau|afk is now known as rpittau | 08:57 | |
| *** ccamacho has joined #openstack-swift | 09:02 | |
| *** nanzha has quit IRC | 09:11 | |
| *** nanzha has joined #openstack-swift | 09:13 | |
| *** nanzha has quit IRC | 09:21 | |
| *** nanzha has joined #openstack-swift | 09:25 | |
| *** csmart has quit IRC | 09:27 | |
| *** baffle has quit IRC | 09:27 | |
| *** openstackstatus has quit IRC | 09:30 | |
| *** nanzha has quit IRC | 09:32 | |
| *** nanzha has joined #openstack-swift | 09:34 | |
| *** nanzha has quit IRC | 09:54 | |
| *** threestrands has joined #openstack-swift | 10:24 | |
| *** rcernin has quit IRC | 10:31 | |
| *** csmart has joined #openstack-swift | 10:36 | |
| *** baffle has joined #openstack-swift | 10:36 | |
| *** rcernin has joined #openstack-swift | 10:46 | |
| *** tkajinam has joined #openstack-swift | 10:50 | |
| *** rcernin has quit IRC | 11:31 | |
| *** threestrands has quit IRC | 11:59 | |
| *** tkajinam has quit IRC | 12:11 | |
| *** tkajinam has joined #openstack-swift | 12:15 | |
| *** tkajinam has quit IRC | 12:16 | |
| *** zaitcev__ has joined #openstack-swift | 13:18 | |
| *** ChanServ sets mode: +v zaitcev__ | 13:18 | |
| *** zaitcev_ has quit IRC | 13:21 | |
| *** psachin has quit IRC | 13:53 | |
| *** mikecmpbll has quit IRC | 15:01 | |
| *** mikecmpbll has joined #openstack-swift | 15:04 | |
| *** hoonetorg has quit IRC | 15:51 | |
| *** ormandj has joined #openstack-swift | 16:00 | |
| ormandj | https://github.com/openstack/swift/blob/f9898ec8f3b27954a6544db91ca36f20b5b74feb/swift/common/middleware/ratelimit.py#L286 <-- any reason 498 vs. 429? | 16:00 |
|---|---|---|
| *** hoonetorg has joined #openstack-swift | 16:04 | |
| *** bharath54321 has joined #openstack-swift | 16:04 | |
| bharath54321 | Hi I have a question regarding Swift. In a GET request, out of the 3 disks(let's say replication factor of 3) which one disk is picked to retrive the object? | 16:05 |
| bharath54321 | Is it random or is there any parameter by which swift picks it? | 16:06 |
| *** mikecmpbll has quit IRC | 16:19 | |
| *** mikecmpbll has joined #openstack-swift | 16:21 | |
| *** diablo_rojo has joined #openstack-swift | 16:36 | |
| root____1 | Hi bharath54321, there is a param for that, you can look for "sorting_method" and "read_affinity" in the docs. The default is shuffle (random) | 16:39 |
| *** root____1 is now known as dcourtoi | 16:40 | |
| bharath54321 | Thanks for the reply! I can't seem to find it in the docs. | 16:41 |
| bharath54321 | Does it have anything to do with the response time of a disk? | 16:41 |
| bharath54321 | An SSD is faster than HDD, will Swift pick an SSD rather than an HDD in this case? | 16:42 |
| dcourtoi | the docs where you can find it : https://docs.openstack.org/swift/ocata/deployment_guide.html and https://docs.openstack.org/swift/pike/overview_global_cluster.html#read-affinity | 16:43 |
| *** zaitcev__ is now known as zaitcev | 16:43 | |
| ormandj | bharath54321: no, there's no implementation of that as far as i've seen | 16:44 |
| bharath54321 | Thanks ormandj and dcourtoi for your replies! | 16:44 |
| bharath54321 | data placement is based on even distribution and unique-as-possible algorithm if I am right? | 16:46 |
| bharath54321 | Do you take into account the speed of the disks to place data in? | 16:46 |
| bharath54321 | i.e SSD over HDD | 16:46 |
| zaitcev | bharath54321: when the builder spreads partitions across devices, it does not have the information about which devices are which. The operator can only include devices into rings, or not. So, you can create a "gold" tier policy that costs more and uses SSDs, and "silver" tier policy that uses HDDs. But the distribution is static and proxy only follows the map that builder created ahead of time. | 16:49 |
| ormandj | bharath54321: i had answered that question with my previous response, sorry for lack of clarity. swift is not preferential in placement based on drive type/speed, as it doesn't currently have insight into that. as zaitcev mentioned there are ways to work around this, but it's not automatic | 16:50 |
| ormandj | swift using a consistent hash ring for placement, which is what the above mentioned ring builder creates | 16:50 |
| zaitcev | ormandj: I think 429 wasn't a standard back when Swift was released. So, since we already had the 499 for timeout, someone used another similar code. | 16:51 |
| zaitcev | ormandj, my cursory web search says that RFC-6585 was only published in 2012. | 16:51 |
| ormandj | zaitcev: fair enough, and i'm guessing no plans to change? found the rfc: https://tools.ietf.org/html/rfc6585 | 16:51 |
| ormandj | hah yeah | 16:51 |
| bharath54321 | zaitcev, Thank you for your kind reply. It is clear to me. Also AFAIK you don't give devices but mount points to the devices. So we don't really know about the low level device | 16:52 |
| bharath54321 | ormandj, Your answer made sense. I just wanted to reinforce it to myself. I apologize for the hassle. | 16:52 |
| ormandj | bharath54321: correct, swift has no insight into the devices themselves currently | 16:52 |
| ormandj | no hassle at all | 16:52 |
| bharath54321 | :) So during a GET request does swift go to all the 3 disks and return 1 of the replicas. Or randomly pick one disk, then if it fails go to the next disk? | 16:55 |
| *** rdejoux has quit IRC | 17:09 | |
| *** openstackgerrit has joined #openstack-swift | 17:21 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Fix BadResponseLength error when creating symlink https://review.opendev.org/696113 | 17:21 |
| zaitcev | Once again, the policy can be shuffle or timing. So, in case of GET, the proxy picks up one node from the list sorted according to the policy. If the connection is established, proxy continues with the GET. | 17:24 |
| bharath54321 | Thanks zaitcev for your time! | 17:24 |
| bharath54321 | It really helped me | 17:25 |
| zaitcev | Helped in what way? What do you want to do with this information? | 17:25 |
| bharath54321 | I am using Swift in Uni for research on power saving algorithms | 17:26 |
| bharath54321 | disk power saving algorithms | 17:27 |
| bharath54321 | I have implemented a non volatile memory cache in Swift to store the data of a sleepig disk | 17:27 |
| bharath54321 | On a controlled setup that is. | 17:27 |
| zaitcev | Curious. | 17:29 |
| bharath54321 | I have managed to sleep a disk in Swift by a lot of hard work :) | 17:29 |
| *** mikecmpbll has quit IRC | 17:36 | |
| tdasilva | bharath54321: that’s really cool. please do share with the community once your research is published | 17:43 |
| bharath54321 | tdasilva, yup, will do :) | 17:48 |
| *** tesseract has quit IRC | 17:53 | |
| *** bharath54321 has quit IRC | 18:11 | |
| *** rpittau is now known as rpittau|afk | 18:14 | |
| *** joeljwright has quit IRC | 18:23 | |
| *** joeljwright has joined #openstack-swift | 18:25 | |
| *** ChanServ sets mode: +v joeljwright | 18:25 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Allow internal clients to use reserved namespace https://review.opendev.org/682138 | 18:27 |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: New Object Versioning mode https://review.opendev.org/682382 | 18:27 |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: s3api: Implement object versioning API https://review.opendev.org/673682 | 18:27 |
| *** spsurya has quit IRC | 19:30 | |
| kota_ | morning | 20:57 |
| kota_ | no meeting? | 21:01 |
| kota_ | hmm | 21:06 |
| kota_ | let me quit and back to my bed today. | 21:06 |
| kota_ | oic https://wiki.openstack.org/wiki/Meetings/Swift clearly says the next is Dec 4. | 21:08 |
| kota_ | make sense | 21:08 |
| kota_ | probably because of thanksgiving | 21:09 |
| zaitcev | Probably. | 21:53 |
| *** rcernin has joined #openstack-swift | 21:56 | |
| mattoliverau | morning | 22:04 |
| mattoliverau | kota_: yeah, timburke said it was cancelled in last weeks meeting because of thanksgiving. | 22:04 |
| *** pcaruana has quit IRC | 22:06 | |
| *** diablo_rojo has quit IRC | 22:08 | |
| *** baojg has quit IRC | 23:32 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!