mlockall & capabilities not working

[jimmidyson]

$ docker run --rm --cap-add=IPC_LOCK --privileged elasticsearch:2.0.0 -Des.bootstrap.mlockall=true

[2015-11-19 20:16:32,368][WARN ][bootstrap                ] Unable to lock JVM Memory: error=12,reason=Cannot allocate memory
[2015-11-19 20:16:32,370][WARN ][bootstrap                ] This can result in part of the JVM being swapped out.
[2015-11-19 20:16:32,370][WARN ][bootstrap                ] Increase RLIMIT_MEMLOCK, soft limit: 65536, hard limit: 65536
[2015-11-19 20:16:32,370][WARN ][bootstrap                ] These can be adjusted by modifying /etc/security/limits.conf, for example: 
        # allow user 'elasticsearch' mlockall
        elasticsearch soft memlock unlimited
        elasticsearch hard memlock unlimited

[jimmidyson]

@pires Any ideas?

[pires]

@jimmidyson works with my image.

$ docker run --rm --cap-add=IPC_LOCK --privileged quay.io/pires/docker-elasticsearch:2.0.0
log4j:WARN No such property [maxBackupIndex] in org.apache.log4j.DailyRollingFileAppender.
log4j:WARN No such property [maxBackupIndex] in org.apache.log4j.DailyRollingFileAppender.
log4j:WARN No such property [maxBackupIndex] in org.apache.log4j.DailyRollingFileAppender.
[2015-11-19 21:10:49,985][INFO ][node                     ] [Impala] version[2.0.0], pid[14], build[de54438/2015-10-22T08:09:48Z]
[2015-11-19 21:10:49,987][INFO ][node                     ] [Impala] initializing ...
[2015-11-19 21:10:50,070][INFO ][plugins                  ] [Impala] loaded [], sites []
[2015-11-19 21:10:50,221][INFO ][env                      ] [Impala] using [1] data paths, mounts [[/data (/dev/sda1)]], net usable_space [15.8gb], net total_space [18.1gb], spins? [possibly], types [ext4]
[2015-11-19 21:10:53,450][INFO ][node                     ] [Impala] initialized
[2015-11-19 21:10:53,462][INFO ][node                     ] [Impala] starting ...
[2015-11-19 21:10:53,627][INFO ][transport                ] [Impala] publish_address {127.0.0.1:9300}, bound_addresses {127.0.0.1:9300}, {[::1]:9300}
[2015-11-19 21:10:53,662][INFO ][discovery                ] [Impala] elasticsearch-default/bOVA1h6jQPejerv0cbClcw
[2015-11-19 21:10:56,768][INFO ][cluster.service          ] [Impala] new_master {Impala}{bOVA1h6jQPejerv0cbClcw}{127.0.0.1}{127.0.0.1:9300}{master=true}, reason: zen-disco-join(elected_as_master, [0] joins received)
[2015-11-19 21:10:56,829][INFO ][http                     ] [Impala] publish_address {127.0.0.1:9200}, bound_addresses {127.0.0.1:9200}, {[::1]:9200}
[2015-11-19 21:10:56,832][INFO ][node                     ] [Impala] started
[2015-11-19 21:10:56,851][INFO ][gateway                  ] [Impala] recovered [0] indices into cluster_state

[jimmidyson]

Raised docker-library/elasticsearch#70

[jimmidyson]

Docker image needs to be run with --ulimit memlock=-1:-1 & then all is good. Now how to do that in Kubernetes?

[pires]

Still open it seems kubernetes/kubernetes#3595

[jimmidyson]

Reading https://www.elastic.co/guide/en/elasticsearch/guide/current/heap-sizing.html looks like we should only resort to enabling mlockall if we can't disable swap or set swappiness to 1. Both these options should be doable for most users so closing this for now.