Seems my RB4011 is crashing/rebooting since being upgraded from 7.6 to 7.7
Nothing much in the logs over than “cloud change time”, sent a supout to support to see if they can see why’s it’s doing it.
Anyone else been having issues with 7.7?
How did you move from ROS6 to ROS7 on that device ?
Upgrade to upgrade to upgrade … ?
It might be time to netinstall.
Nope…
How fast after boot does it happen?
Have you sent the supout file yet?
Anything exotic in your config?
Seems random, yes I sent a supout.
Config has a L2TP tunnel and a second routing table to use it/ some routing rules but otherwise no not particularly exotic.
Take in consideration what @holvoetn mentioned: export your complete config (/export show-sensitive file=anynameyoulike), do a netinstall and perform an import of the config (/import file=anynameyoulike). Hope it helps!
I noticed a memory leak in the DNS server in version 7.7 on my RB4011. Has already been reported to support, have not received an answer so far.
Maybe this is the cause of some crashes.
I noticed that on all mikrotik devices including CRS (where CPU doesn’t really do anything besides SNMP regularly). Also virtual machines on x86 have the same issue.
Forgot to update this thread:
Had a support ticket open about it and they kept wanting me to try a different power source, even after I’d discovered that removing a un-needed bridge configuration stabilised it and told them this.
I had eth1 and eth2 in a bridge but as nothing was connected to eth2 anymore it was un-needed legacy configuration, since removing the bridge it’s been stable
I have a memory leak also on multiple routers that run as Hotspot controllers. I have simplified the bridging down to the lowest point where it will still function and will see how it goes.
I went to 7.8 and the memory leaks seem to have stablised considerably. Not entirely convinced they are gone, but it is clear they are significantly improved.
Also noticed same on a hAP ac lite unit for which the upgrade to 7.8 completely fixed it.
7.8 has indeed resolved the issue. Although nothing in the changelogs to indicate this should be the case or if something was actually identified that I can find.