Created attachment 16178 [details]
We are receiving the following PANIC on our samba domain controllers:
smb_panic_default: PANIC (pid 2024536): Bad talloc magic value - unknown value
Please see attached file for log excerpt which includes the full backtrace.
This has been occurring for over a year, but has increased in frequency to the point where it is causing us operational problems. My guess as to the reason for the increase in frequency is increased load on the domain controllers owing to additional staff. The DC which exhibits the issue most often is running at our hub site, has the most users and the additional load of RADIUS wireless authentication for all sites and an on site IIS server also authenticating users at all sites.
The PANIC is at present occurring approximately once per day, though it sometimes becomes more frequent up to every five minutes for about an hour.
Machines attempting to authenticate when the PANIC occurs do not appear to realize the process they were talking to has died, and take a few minutes to restart the authentication.
The issue has presented itself on:
* CentOS 6.10 with samba 4.10.10 compiled from source
* CentOS 8.1 with samba 4.12.5 and samba 4.9.18 both compiled from source
* Ubuntu 20.04.1 with samba 4.11.6 from the Ubuntu repositories and samba 4.12.6 from the apt.van-belle.nl repository
When compiling from source the following configure line is used:
./configure --with-ads --with-shared-modules=idmap_ad
Six samba domain controllers, one at each physical site. All are ESXi 6.0 virtual machines. Three are currently running CentOS 6.10, two Centos 7.7. Those five are all running samba 4.10.10 compiled from source.
There are no Windows domain controllers though there are Windows 2012 R2 domain members.
(The logs of four of those six DCs currently contain the panic, the two which do not are one of the original CentOS 6 and one of the CentOS 7 VMs, though it may have occurred there but rotated out of the logs).