Database gone after either upgrade or OOM

#1

Previously we had an old v0.13 version of influxDB running. I shutdown the server, moved the data folder to a backup location, and installed influxDB 1.3.1. Things were running great for about 2 weeks. All of a sudden, influx went down due to an OOM issue. To bring it back online, I had to make a few changes to the config file (removing references to clustering, and adding a second set of [] tags surrounding the collectd and opentsdb sections. Both collectd and opentsdb have enabled=false)

When the server came back online, the database I had been using was missing. I checked on disk, and it had also removed the files from disk. I don’t know if this happened with the config file was changed, or when the OOM condition happened. I find it odd that either case would physically delete the meta and data files from disk.

Any ideas what caused this? We are looking to move our production boxes to the new version, but am hesitating to do so with this happening.

#2

@pauldix this is the issue I referred to you on Twitter. It may or may not be related to the Boris’s issue but this has been seen.

#3

The actual files in the data and meta directories were removed? The database only removes files from data in the case of dropping a database or if you have a retention policy that expires data out it will drop old shards. I wouldn’t expect data to just be removed and haven’t heard this kind of report before. Any other possibilities?