Intermittent watchdog timeouts

[Dwayne Rightler]

Hello, we've been trying to troubleshoot why we're getting stuck threads which trigger the watchdog abort.  It's intermittent and we're kind of stumped at this point.  If I provide a coredump and the binary could I get some assistance in determining the potential cause of the issue?  Or would this be better placed under Issues on github?

Thanks,

Dwayne

[René Cannaò]

Hi Dwayne,

Thank you for using ProxySQL !

You mentioned that the issue is intermittent: I need to get more details. Does this mean that it starts logging errors about missing heartbeat, but does not always lead to an abort?

If this is the case, there is a very high probability that you are suffering from resource constraints (CPU, RAM, network).

The most common root causes are:

* aggressive swap usage

* CPU being throttled (for example due to Kubernetes settings, or burst cloud instances)

Thanks,

René

--
You received this message because you are subscribed to the Google Groups "proxysql" group.
To unsubscribe from this group and stop receiving emails from it, send an email to proxysql+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/proxysql/292341af-fb03-4cc4-950a-6d05bc754a64n%40googlegroups.com.

[Dwayne Rightler]

Hi René,

Thanks for your quick reply.

Once we start logging missed heartbeats, it always leads to an abort. 

Instead of intermittent, perhaps I should have said unpredictable. We're currently running load tests in our test lifecycle, and the watchdog errors don't appear in every load test, or at consistent points in the load test when they do occur.  It happens with and without idle-threads enabled. Both memory and cpu usage on the proxysql VMs are under 50%.

Thanks,
Dwayne