LSWS 6.3.3 Sever unexpectedly dropping connections

MarkD

We are now running LSWS 6.3.3 build 0 and we have been getting reports from end users of dropped connections to the servers. Nothing shows up in the security_audit.log file so I don't think it's a firewall issue but clearly something has changed from 6.3.2 because if we roll back the problem goes away but then we are faced with random processes being killed 🤦‍♂️

If anybody else is seeing this I would love to know - I have raised this already with Litespeed support and will keep you updated

cPFence

MarkD because if we roll back the problem goes away but then we are faced with random processes being killed 🤦‍♂️

Are you also seeing random processes being killed on the older stable versions like LSWS 6.2.2? Lately, running the latest versions of both LSWS and OLS seems to carry more risk than benefit.

AdamM

I have warns on one of the servers:
[3436959] [T0] [AutoRestarter] forking too frequently, suspend for a while!

Also this...
At [03/Jun/2025:17:02:30 +0000], web server with pid=3454591 received unexpected signal=11, a core file has been created. A new instance of web server will be started automatically!

Did you have that too?

MarkD

Yes, we are seeing the AutoRestarter logged whenever the problem happens and we have also just realised that segfaults are being logged for litespeed in the syslog.

2025-06-04T14:21:09.361865+01:00 vps539 kernel: litespeed[46499]: segfault at 20 ip 0000000000825fe8 sp 00007fff4fb0b4a0 error 4 likely on CPU 1 (core 0, socket 0)
2025-06-04T14:23:47.722910+01:00 vps539 kernel: litespeed[47036]: segfault at 20 ip 0000000000825fe8 sp 00007fff4fb0b4a0 error 4 likely on CPU 0 (core 0, socket 0)

All information has been sent to litespeed support and I am waiting for a response...

MarkD

cPFence We haven't tried rolling back to that version yet as my understanding is that this has always been a problem but when it was running under docker litespeed had no access to kill other processes like mysql and redis so it was never detected.

I assume if we roll back to any version of LS the problem will still be there for v12 users like us.

MarkD

I have just downgraded one of the servers to 6.0.12, lets see how it runs overnight as its the end of the day here in the UK.

MarkD

Litespeed support updated one of our test servers to their latest debug build of 6.3.3 and so far we have been unable to make that server crash!

We have now rolled that version out to all of our servers and we are monitoring the situation by looking for any core dump files across all servers.

If you want to check if your litespeed version has had any issues have a look in this folder for any litespeed core dump files

ls -l /var/lib/apport/coredump

If we can get to the end of the day with no core dump files being created anywhere then I think we can say that is finally fixed! 🥳

twest

MarkD thank you for your service.

MarkD

And we are done - 6.3.3 build 1 is now available from the release channel

/usr/local/lsws/admin/misc/lsup.sh -f -v 6.3.3 -b 1