Memory leak in XDATA on 13.1

[TheWanderer_360]

So we got a pair of 10200s with 48GB of RAM running on 13.1.0.3 (same results in 13.0) with only thing provisioned is LTM. TMM memory usage keeps climbing until it hits 85% and the aggressive sweeper mode kicks in killing everything in sight. TMM starts at 10% after the reboot and climbs to 85% within 3 weeks, the TMM memory graph just goes up (never releases RAM)

show sys memory

shows xdata subsystem eating up 32gigs before we reboot. Other memory is solid at 40%, SWAP is 0% and CPU usage is never above 15%. Any way to find out what's XDATA is holding on to? sending qkview to support is not an option due to environment, perhaps we can parse it ourselves?

[Simon_Blakely]

It is going to be hard to resolve this if you cannot provide data (QKView, tmm core) to Support for identification of the memory leak.

What sort of profiles do you use on your virtuals?

[TheWanderer_360]

Thanks for your reply. Yeah we've been battling with this for some time now. Typically we just use http, client-side ssl, or just fastl4

[Simon_Blakely]

You say typically. Could someone have added a mtqq profile (it's a new feature in 13) to a virtual?

There are two possibilities - you are hitting something already known (in which case it may be easy identify from the config). However, the list of known tmm memory leaks into XDATA is quite short list.

Alternatively, you are hitting something completely new. In that case, Development need the core file and QKView to identify and isolate the issue.

Without providing data to F5 Support, all you can do is keep trying upgrades until the issue is fixed. If it is an unknown issue, it may never be fixed, because no-one else may have the same combination of traffic and config.