I'm currently running into a problem where the watch-history record for a particular watcher isn't index due complains about the fieldname ctx.payload.hits.total:
[2017-01-31 11:33:51,687][ERROR][watcher.execution ] failed to update watch record [recovery_ceased_log_source_***_dev_wa_userapi_cus_1689-2017-01-31T11:33:51.268Z] MapperParsingException[Field name [ctx.payload.hits.total] cannot contain '.']
This is unfortunate, as I need my watch-history-* indices to contain 100% accurate data. I have a series of "recovery" watchers that read through the execution history, and send a 'recovered' message when things are back to normal (e.g excess disk usage watcher stopped triggering)
Is there any easy way to fix this?
The watcher that triggered this error is included below. It is a chained-input watcher that checks:
did the ceased_log_source watcher execute?
did the ceased_log_source watcher "recover"?
did this watcher trigger?
if so, don't pass this watcher's condition
otherwise, was the input-order the order executed -> recovered?
After a quick glance I think there is an issue with the mapping of chained inputs, where fields are accidentally mapped that should not be... will keep you posted
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.