Getting a lot of "nothing needed to be retained" INFO

[John Wong <go...@gmail.com>]

Hi

I am on Cassandra 1.2.19 and we are running OpsCenter 5.1.4 (and agent
5.1.4).

We recently started to observe high load occurring in our dev cluster.
There was some small spikes in terms of amount of requests going into the
cluster so we are not ruling out that possibility.

What we see is we get memtable filling up, and some gc kicks in. So the
load is likely contributed by gc. However, we are puzzled why the following
INFO logs are constantly appearing in our log. We are not sure the meanings.

Do we need to worry about these INFO? We suppress our high load by shutting
down datastax-agent running on each node. That helped. My understanding is
the agent adds more load because of the metrics collection.

INFO 21:42:36,456 Completed flushing; nothing needed to be retained.
Commitlog position was ReplayPosition(segmentId=1447227203262,
position=9710385)
 INFO 21:43:36,458 Enqueuing flush of Memtable-batchlog@1532343528(12360/38070
serialized/live bytes, 120 ops)
 INFO 21:43:36,459 Writing Memtable-batchlog@1532343528(12360/38070
serialized/live bytes, 120 ops)
 INFO 21:43:36,463 Completed flushing; nothing needed to be retained.
Commitlog position was ReplayPosition(segmentId=1447227203262,
position=9795055)

Thanks.

John

Re: Getting a lot of "nothing needed to be retained" INFO

[Michael Shuler <mi...@pbandjelly.org>]

On 12/03/2015 03:49 PM, John Wong wrote:
> Hi
> 
> I am on Cassandra 1.2.19 and we are running OpsCenter 5.1.4 (and agent
> 5.1.4).
> 
> We recently started to observe high load occurring in our dev cluster.
> There was some small spikes in terms of amount of requests going into
> the cluster so we are not ruling out that possibility.
> 
> What we see is we get memtable filling up, and some gc kicks in. So the
> load is likely contributed by gc. However, we are puzzled why the
> following INFO logs are constantly appearing in our log. We are not sure
> the meanings.
> 
> Do we need to worry about these INFO? We suppress our high load by
> shutting down datastax-agent running on each node. That helped. My
> understanding is the agent adds more load because of the metrics collection.
> 
> INFO 21:42:36,456 Completed flushing; nothing needed to be retained. 
> Commitlog position was ReplayPosition(segmentId=1447227203262,
> position=9710385)
>  INFO 21:43:36,458 Enqueuing flush of
> Memtable-batchlog@1532343528(12360/38070 serialized/live bytes, 120 ops)
>  INFO 21:43:36,459 Writing Memtable-batchlog@1532343528(12360/38070
> serialized/live bytes, 120 ops)
>  INFO 21:43:36,463 Completed flushing; nothing needed to be retained. 
> Commitlog position was ReplayPosition(segmentId=1447227203262,
> position=9795055)

INFO log lines are informational level entries of the actions the node
is performing.

One can configure logging for WARN (and higher priority) logs, if the
INFO lines are not desired.

In conf/log4j-server.properties change the rootLogger from INFO to WARN:

-log4j.rootLogger=INFO,stdout,R
+log4j.rootLogger=WARN,stdout,R

-- 
Kind regards,
Michael