632 reputation
11336
bio website craters-of-the-mind.blogspot.…
location Moscow, Russia
age 33
visits member for 5 years, 7 months
seen Nov 18 at 16:26

:-)

mailto: ag@logiceditor.com


Nov
6
comment Uneven CPU core utilization
Testing is done by running ab -n 100000 -q -c 20 http://my.example.com, from six different external machines.
Nov
6
revised Uneven CPU core utilization
clarified
Nov
6
comment Uneven CPU core utilization
Worker processes are Lua-based, and do some non-trivial business logic (including communication over network with other machines), so 800 rps is more or less OK (but there is some room for optimization). That being said, please note that this question is specifically about uneven CPU core utilization.
Nov
6
comment Uneven CPU core utilization
Well, all cores except the first one are consistently under-loaded, and first core is consistently over-loaded. Looks fishy to me.
Nov
6
revised Uneven CPU core utilization
clarified subject
Nov
6
asked Uneven CPU core utilization
Nov
3
awarded  Nice Question
Oct
30
comment Troubleshooting poor disk IO performance
deadline on Xen, noop on VM seems to be the best option. We still have to analyze data properly, but while other combination trigger 500-1100 502 errors (with stable traffic load), while deadline-noop triggers 0-600 502s (with the same load). However it still stalls. Is there anything else we can tune?
Oct
30
comment Troubleshooting poor disk IO performance
No battery-backed or flash-backed cache, sadly. This is a plain and humble Hetzner box.
Oct
30
comment Troubleshooting poor disk IO performance
It seems that I'm using deadline scheduler. Will try noop and (for completeness) cfq.
Oct
30
comment Troubleshooting poor disk IO performance
See dstat and iostat dumps here: groups.google.com/forum/#!msg/redis-db/vSAvnYVtX9w/NRYlBwOrCTsJ
Oct
30
comment Troubleshooting poor disk IO performance
It says noop [deadline] cfq on VM.
Oct
30
comment Craft a file that ignores all writes without being marked as read-only
Interesting idea, thank you. Looks like a way to go, unless there is a ready-made solution of some kind.
Oct
30
comment Craft a file that ignores all writes without being marked as read-only
Unfortunately, no. The whole point of this feat is to ensure that if the program (actually a continuously running service) restarts, it has actual data to load.
Oct
30
revised Craft a file that ignores all writes without being marked as read-only
clarified why can't I use /dev/null
Oct
30
comment Craft a file that ignores all writes without being marked as read-only
Looks not robust enough...
Oct
30
asked Craft a file that ignores all writes without being marked as read-only
Oct
30
accepted s3fs: how to force remount on errors?
Oct
30
answered s3fs: how to force remount on errors?
Oct
30
revised Troubleshooting poor disk IO performance
not so Redis-specific as it might appear from a first glance