Out of Memory (Fatal) Error when Dropping Measurement

[goakley]

Bug report

System info:
Debian GNU/Linux 8 (jessie)
3.16.0-4-amd64 #1 SMP Debian 3.16.7-ckt25-1 (2016-03-06) x86_64 GNU/Linux
InfluxDB 0.13.0

Steps to reproduce:

1. Collect around 50,000 data points in a single measurement
2. Attempt to drop the measurement with DROP MEASUREMENT [name]
3. Wait a few seconds

Expected behavior:
After some time, InfluxDB should signify that it has successfully dropped the measurement, or that it has encountered some error and cannot finish dropping the measurement.

Actual behavior:
InfluxDB crashed and took a few minutes to restart from WAL segments. Running DROP MEASUREMENT [name] a second time succeeded.

Additional info: [Include gist of relevant config, logs, etc.]
Log contents (includes stack trace)

[joelegasse]

@goakley Can you see if @jwilder's change fixes your issue?

[jwilder]

Going to close this as it should be fixed by #7015. The logs show hundreds of writes blocked on the tsdb.Store mutex. DropMeasurement incorrectly was holding the writing lock which locks up the whole tsdb.Store for writes and queries. The lock caused writes to pile up and ultimately a compaction ran and allocation triggered an OOM.