Memory Leak in ROS 6.11

Memory utilization on MikroTik 1100AHx2 acting as the PPTP and SSTP server to ~ 170 MikroTik RB443 and 951G systems, each acting as a VPN client continues to climb. After reboot and all VPNs are connected, memory utilization is about 41MB. After couple of weeks memory utilization has risen to 290+ MB. Traffic through router is quite constant. sooner or later, ROS will disconnect all VPNs, reducing memory requirement, but now down to initial requirement. after which memory continues to climb again.
Rebooting router reduces memory utilization back down to 41MB.

Anyone else experiencing ROS memory leak with VPNs?
Memory.jpg

Now 6.12 is out,

test 6.12 and report to support@mikrotik.com if the same problem are present.

Already reported to support as [Ticket#2014050766000641]. However no response from Mikrotik.
Failing system is in production and updates need to be scheduled during off hours.
Since the issue takes several days to demonstrate failure, scheduling an outage to upgrade to a new version that has no expectation of success is difficult to justify.
Can you point me to release notes indicating 6.12 has a fix for this problem? We believe root cause may be associated with the unreliability of SSTP (and PPTP) connections where they will occasionally connect and disconnect in a 2-3 minute cycle.

I use RB1100AHx2 for works,
upgrade to 6.12 at @ 04:00 AM has required only 1 min…

1 minute if and only if it is successful. Otherwise much longer time for netinstall, testing, restoring backups and so on. At critical installations I suggest to leave actual router with its settings and just unplug it. Put other device upgraded instead of it. If whatever fails the connecting otiginal device back is really just matter of 1 minute.

+1 In my experience, no upgrade, no matter how minor, ever takes one minute.
Especially when ROS upgrades are not guaranteed to be 100% backwards compatible.

While not VPN specific, we came across a memory leak similar to what you are seeing in large hotspot deployments (20,000+ users). You could watch the RAM drain over a period of time and then the box would lock and reboot. This was in the 5.2x x86 train and we moved to CCRs which improved stability and we didn’t experience it after that

FYI: have updated to both 6.12 and then to 6.13. Problem still exists on both ROS versions