7.10.0 elasticsearch period crash by stackoverflow

hello..recentlly I got a weird elasticsearch crash problem and it lasts nearly one month .since I reindex some of my index to enlarge the shards from 1 to 6..and the error like following

[o.e.b.ElasticsearchUncaughtExceptionHandler] [es-cn-9lb3e7861000eg3q5-7819d1df-0001] fatal error in thread [elasticsearch[es-cn-9lb3e7861000eg3q5-7819d1df-0001][clusterApplierService#updateTask][T#1]], exiting
java.lang.StackOverflowError: null
at org.elasticsearch.index.mapper.ParametrizedFieldMapper$Parameter.getDefaultValue(ParametrizedFieldMapper.java:214) ~[elasticsearch-7.10.0.jar:7.10.0]
at org.elasticsearch.index.mapper.ParametrizedFieldMapper$Parameter.getDefaultValue(ParametrizedFieldMapper.java:214) ~[elasticsearch-7.10.0.jar:7.10.0]
at org.elasticsearch.index.mapper.ParametrizedFieldMapper$Parameter.getDefaultValue(ParametrizedFieldMapper.java:214) ~[elasticsearch-7.10.0.jar:7.10.0]
at org.elasticsearch.index.mapper.ParametrizedFieldMapper$Parameter.getDefaultValue(ParametrizedFieldMapper.java:214) ~[elasticsearch-7.10.0.jar:7.10.0]
at org.elasticsearch.index.mapper.ParametrizedFieldMapper$Parameter.getDefaultValue(ParametrizedFieldMapper.java:214) ~[elasticsearch-7.10.0.jar:7.10.0]
at org.elasticsearch.index.mapper.ParametrizedFieldMapper$Parameter.getDefaultValue(ParametrizedFieldMapper.java:214) ~[elasticsearch-7.10.0.jar:7.10.0]

then I search the source code with elasticsearch 7.10.0 ..find the Parameter.getDefaultValue canbe infinite loop when it is the Parameter<NamedAnalyzer> ..but don't in which case it can be triggered.
this crash periodlly happens ,in 4:00 10:00 16:00 22:00 ..and every master node crashs one by one ..then the cluster back to normal ..

any idea about this looking forword your replies thanks~

I'd upgrade to 7.17 at least. And better is to switch to Elasticsearch 8.14.2.

1 Like

Thanks a lot for replying!!finally I get the exception stack from the /_nodes/hot_threads found a period index updating cause the merge error, but don't know why crash every 6 hours instead of crash immediately :smiling_face_with_tear: , whatever currently we can avoide this error , and may upgrade the version in further

1 Like