Where the filter cache memory allocated?

Hi Team,

the filter cache memory is allocated from heap?

if it is allocated from heap, is the possibility to affect the gc?

It is taken from heap but it isn't GCd. So when something goes into filter cache, it takes away heap for other things.

It shouldn't be a major issue though, the bitsets are very efficient.

Hi @warkolm,
Thanks for you reply, we are facing gc issue on prod servers.

  1. we are using elasticsearch 1.3.7
  2. we allocated 10G for each clusters
  3. filter and field cache 30%, 20%
  4. gc params like below
    -XX:+UseConcMarkSweepGC -XX:+UseParNewGC -XX:CMSInitiatingOccupancyFraction=65 -XX:+UseCMSInitiatingOccupancyOnly -XX:+PrintGCDetails -verbose:gc -XX:+PrintGCDateStamps -XX:NewSize=1536m -XX:MaxNewSize=1536m -XX:SurvivorRatio=46 -XX:MaxTenuringThreshold=3

we are getting promotion failed , and concurrent mode failure gc issues. ( nearly 7 to 8 G heap usage )
once we cleared the cache the alert gone.

can you give me suggestions...

Upgrading would be my first recommendation, 1.3 hasn't been supported for a loooooong time.

But i need to know, why "the promotion failed , and concurrent mode failure" alerts gone once i cleared cache.