Collector crashes

If you find the collector crashing, the most likely cause is running out of memory. We recommend a small instance size by default as that is enough for the typical case, but with a very large schema or a large or diverse query workload, that may not be sufficient.

Collector memory may be limited either through available instance memory, or through systemd (if your platform uses it).

If you believe the collector is crashing for other reasons, please contact us.

Systemd limits

To avoid impacting other programs on the same instance, the pganalyze collector runs with a memory limit of 1024MB by default.

If running out of memory, you may see log messages like the following:

Dec 31 06:20:17 ip-1.compute.internal pganalyze-collector[5568]: I [server1] Submitted full snapshot successfully
Dec 31 06:20:25 ip-1.compute.internal systemd[1]: pganalyze-collector.service: main process exited, code=killed, status=9/KILL
Dec 31 06:20:25 ip-1.compute.internal systemd[1]: Unit pganalyze-collector.service entered failed state.
Dec 31 06:20:25 ip-1.compute.internal systemd[1]: pganalyze-collector.service failed.

The best way to resolve that is to define a local override for the limit. Run sudo systemctl edit pganalyze-collector.service and then enter

[Service]
MemoryLimit=2048M

(or whatever limit you feel is appropriate) in that file. Then restart the collector with sudo systemctl restart pganalyze-collector.

Instance limits

If you are not using systemd or you have increased those limits to available instance memory and you are still seeing problems, try running the collector on a larger instance. If that does not help, please contact us.