addthis/hydra

excessive zk logging on empty/idle cluster

stewartoallen opened this issue · 3 comments

i've started and stopped a virgin local stack a few times then and after letting it sit idle for 15-20 minutes, randomly did a du -sh on the hydra run directory. surprisingly it came it at close to 500MB. i drilled down and found a huge chunk of zookeeper logs in hydra/etc/zookeeper/version-2/ ... it seems they are all /spawn/queue updates. not clear why that would be given the fact that the cluster was completely idle and lacked jobs, tasks, hosts, etc. not sure how to really debug this or if i want to (i don't).

I think you are sure that you do not want to debug :). We will investigate

On Friday, February 21, 2014, Stewart Allen notifications@github.com
wrote:

i've started and stopped a virgin local stack a few times then and after
letting it sit idle for 15-20 minutes, randomly did a du -sh on the hydra
run directory. surprisingly it came it at close to 500MB. i drilled down
and found a huge chunk of zookeeper logs in hydra/etc/zookeeper/version-2/
... it seems they are all /spawn/queue updates. not clear why that would be
given the fact that the cluster was completely idle and lacked jobs, tasks,
hosts, etc. not sure how to really debug this or if i want to (i don't).

Reply to this email directly or view it on GitHubhttps://github.com//issues/34
.

I was not able to repeat this. I created a fresh vagrant box. Using
the commands from this gist:

https://gist.github.com/abramsm/9176182

I let it run for hours and the total local stack size was 1.1MB. You
mentioned that you have zero hosts, so it sounds like you are not
using our standard local stack. Can you explain more about your setup
and changes from master? Directory contents below.

vagrant@precise32:~/hydra/hydra-local$ du -kh

8.0K ./bin

808K ./zoo/version-2

812K ./zoo

4.0K ./log/qmaccess

16K ./log/spawn-stats/2014/2/20

20K ./log/spawn-stats/2014/2

24K ./log/spawn-stats/2014

28K ./log/spawn-stats

4.0K ./log/events-query

4.0K ./log/events-trace

4.0K ./log/events-worker

136K ./log

4.0K ./streams

8.0K ./mini2n

8.0K ./mini0n

4.0K ./etc/jobstore/.git/logs/refs/heads

8.0K ./etc/jobstore/.git/logs/refs

12K ./etc/jobstore/.git/logs

4.0K ./etc/jobstore/.git/objects/pack

4.0K ./etc/jobstore/.git/objects/info

12K ./etc/jobstore/.git/objects

4.0K ./etc/jobstore/.git/refs/tags

4.0K ./etc/jobstore/.git/refs/heads

12K ./etc/jobstore/.git/refs

4.0K ./etc/jobstore/.git/branches

4.0K ./etc/jobstore/.git/hooks

56K ./etc/jobstore/.git

4.0K ./etc/jobstore/jobs

64K ./etc/jobstore

76K ./etc

36K ./pid

8.0K ./mini1n

4.0K ./query.tmpdir

4.0K ./exec

1.1M .

On Fri, Feb 21, 2014 at 5:34 PM, Matt Abrams abramsm@gmail.com wrote:

I think you are sure that you do not want to debug :). We will investigate

On Friday, February 21, 2014, Stewart Allen notifications@github.com
wrote:

i've started and stopped a virgin local stack a few times then and after
letting it sit idle for 15-20 minutes, randomly did a du -sh on the hydra
run directory. surprisingly it came it at close to 500MB. i drilled down and
found a huge chunk of zookeeper logs in hydra/etc/zookeeper/version-2/ ...
it seems they are all /spawn/queue updates. not clear why that would be
given the fact that the cluster was completely idle and lacked jobs, tasks,
hosts, etc. not sure how to really debug this or if i want to (i don't).

Reply to this email directly or view it on GitHub.

Closed due to inactivity