How often should compaction run after 24 hours and how long should it take?

We are running 1.0.0-beta1 with just over 1 million series. Server has 16 cores 
and 32GB of RAM. Writing less than 5000 points per second.

For the first 24 hours, query response time is less than a second. Level 1/2 
compaction takes place every 20 minutes and completes in seconds. Everything is 
fast.

Then, after the 24 hours configured in "compact-full-write-cold-duration", full 
compaction kicks in and runs continuously and performance takes a noticeable 
hit. Queries can take several seconds to complete.

The compaction appears to be ongoing after 24 hours - running every minute for 
more than a minute and constantly writing at approx 100MB/s to disk (RAID array 
not SSD).

Is this normal?

If so, are there config parameters that can be tweaked? or will a move to SSDs 
resolve the problem?

Here's an example where the influx process was restarted at about 11:20am on 
2016/06/14 - after 24 hours, you can see in the log file the switch from 
infrequent level 1/2 to ongoing full compaction before and after 11:20am.

% grep "files into" influxdb.log
[tsm1] 2016/06/15 09:10:24 compacted level 1 group 0 of 2 files into 1 files in 
12.953107399s
[tsm1] 2016/06/15 09:10:44 compacted level 2 group 0 of 2 files into 1 files in 
20.321327593s
[tsm1] 2016/06/15 09:28:29 compacted level 1 group 0 of 2 files into 1 files in 
14.123922529s
[tsm1] 2016/06/15 09:47:39 compacted level 1 group 0 of 2 files into 1 files in 
13.128642782s
[tsm1] 2016/06/15 09:47:57 compacted level 2 group 0 of 2 files into 1 files in 
17.759666145s
[tsm1] 2016/06/15 10:06:34 compacted level 1 group 0 of 2 files into 1 files in 
11.840285989s
[tsm1] 2016/06/15 10:26:02 compacted level 1 group 0 of 2 files into 1 files in 
12.742921103s
[tsm1] 2016/06/15 10:26:21 compacted level 2 group 0 of 2 files into 1 files in 
18.407723192s
[tsm1] 2016/06/15 10:43:52 compacted level 1 group 0 of 2 files into 1 files in 
11.848901735s
[tsm1] 2016/06/15 11:03:10 compacted level 1 group 0 of 2 files into 1 files in 
13.017833455s
[tsm1] 2016/06/15 11:03:28 compacted level 2 group 0 of 2 files into 1 files in 
18.295207874s
[tsm1] 2016/06/15 11:05:06 compacted level 3 group 0 of 4 files into 1 files in 
1m36.54105056s
[tsm1] 2016/06/15 11:06:19 compacted full 2 files into 1 files in 
1m13.027526301s
[tsm1] 2016/06/15 11:22:07 compacted level 1 group 0 of 2 files into 1 files in 
12.6243752s
[tsm1] 2016/06/15 11:30:51 compacted full 2 files into 1 files in 479.64141ms
[tsm1] 2016/06/15 11:32:42 compacted full 2 files into 2 files in 
1m52.257509758s
[tsm1] 2016/06/15 11:32:55 compacted full 2 files into 2 files in 2m4.496843095s
[tsm1] 2016/06/15 11:33:53 compacted full 2 files into 2 files in 1m9.375094851s
[tsm1] 2016/06/15 11:34:00 compacted full 2 files into 2 files in 1m3.399309336s
[tsm1] 2016/06/15 11:34:56 compacted full 2 files into 2 files in 1m2.215633678s
[tsm1] 2016/06/15 11:35:08 compacted full 2 files into 2 files in 1m7.2966692s
[tsm1] 2016/06/15 11:36:00 compacted full 2 files into 2 files in 1m2.651181158s
[tsm1] 2016/06/15 11:36:15 compacted full 2 files into 2 files in 1m6.527896164s
[tsm1] 2016/06/15 11:37:05 compacted full 2 files into 2 files in 1m4.387296019s
[tsm1] 2016/06/15 11:37:31 compacted full 2 files into 2 files in 
1m14.593487299s

-- 
Remember to include the InfluxDB version number with all issue reports
--- 
You received this message because you are subscribed to the Google Groups 
"InfluxDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/influxdb.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/influxdb/5662f7cc-7514-4105-b175-eb110da6dc74%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to