openstackgerrit | Merged openstack/swift master: tests/py3: Improve header casing https://review.opendev.org/678674 | 00:09 |
---|---|---|
*** rcernin has quit IRC | 00:30 | |
*** rcernin has joined #openstack-swift | 00:31 | |
*** rcernin has quit IRC | 02:27 | |
*** rcernin has joined #openstack-swift | 02:27 | |
*** zaitcev has quit IRC | 05:57 | |
*** trungnv has quit IRC | 06:01 | |
openstackgerrit | Merged openstack/python-swiftclient master: Update master for stable/train https://review.opendev.org/682166 | 08:09 |
openstackgerrit | Merged openstack/python-swiftclient stable/train: Update .gitreview for stable/train https://review.opendev.org/682165 | 08:09 |
*** rcernin has quit IRC | 08:49 | |
*** e0ne has joined #openstack-swift | 11:51 | |
openstackgerrit | Nguyen Quoc Viet proposed openstack/swift master: versioned_writes checks for SLO object before copy https://review.opendev.org/681656 | 11:54 |
*** e0ne has quit IRC | 13:07 | |
*** pcaruana has joined #openstack-swift | 14:42 | |
donnyd | So I am storing system backups and log files in swift and I was thinking about moving to erasure coding. Will it work for that use case | 15:34 |
donnyd | I will have 8 swift storage nodes | 15:34 |
donnyd | each with about 2TB of disk | 15:35 |
donnyd | Its nvme ($$$), so I am trying to maximize usable space... but for obvious reasons I think integrity and availability need to come first | 15:36 |
donnyd | Any ideas on erasure coding? Is anyone out there using it in prod? | 15:36 |
*** baojg has quit IRC | 16:43 | |
*** baojg has joined #openstack-swift | 16:43 | |
*** baojg has quit IRC | 16:44 | |
*** baojg has joined #openstack-swift | 16:44 | |
*** baojg has quit IRC | 16:44 | |
*** baojg has joined #openstack-swift | 16:46 | |
*** baojg has quit IRC | 16:46 | |
*** baojg has joined #openstack-swift | 16:50 | |
*** baojg has quit IRC | 16:51 | |
*** baojg has joined #openstack-swift | 16:52 | |
*** baojg has quit IRC | 16:52 | |
*** baojg has joined #openstack-swift | 16:54 | |
*** baojg has quit IRC | 16:55 | |
*** baojg has joined #openstack-swift | 16:58 | |
*** baojg has quit IRC | 16:59 | |
DHE | donnyd: not gone live yet, but I experimented with it on my 30-HDD lab. you definitely want to install a 3rd party package like intel's ISA-L package for performance reasons. but my intel core i7 4xxx series CPU handles gigabit speeds no problem with it. | 17:02 |
DHE | still, why nvme? | 17:03 |
donnyd | Well the spinners make lots of heat | 17:03 |
donnyd | and its what everything else is built on | 17:03 |
donnyd | nova, glance, cinder are currently using all nvme | 17:04 |
donnyd | This isn't really a performance thing, its more about making complete use of the nvme I have in nova | 17:04 |
donnyd | there is 3.2Tb of nvme in each nova server and I use about 500G | 17:05 |
donnyd | so I plan to move swift into vm's and make use of the other 2.5TB per compute | 17:05 |
donnyd | or close to 20Tb of storage | 17:05 |
donnyd | Its mostly about saving the 8 degrees of heat that the spinners make | 17:06 |
donnyd | Also the slowest link i have is 10G... most of it is 40 | 17:07 |
donnyd | so ISA-L will help with the erasure coding part DHE | 17:23 |
openstackgerrit | Merged openstack/swift master: Remove unneeded Zuul branch matcher https://review.opendev.org/681981 | 17:29 |
DHE | donnyd: the stock liberasure algorithm is slow, but functional. whereas isa-l was written by intel and supports AVX optimizations, etc. | 18:01 |
DHE | but you need to pick a mechanism and stick with it | 18:01 |
donnyd | DHE: that is good information to know. | 18:02 |
donnyd | Thank you for helping (again) | 18:02 |
DHE | just in case I should clarify. you still need to install liberasure. it will load isa-l and use that for its heavy lifting though | 18:40 |
*** mvkr has quit IRC | 20:34 | |
*** mvkr has joined #openstack-swift | 20:47 | |
*** pcaruana has quit IRC | 21:24 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!