ssh crash explained

ssh crash explained

[Christopher Faylor]

In short:  I did it.

I scan the system on a regular basis looking for stuck ports using a
utility I wrote called "portusage".  It appeared that this perl script
was running fine on the new system but the output was sufficiently different
that I ended up killing what I thought was a stuck ssh process but what was,
in fact, the master ssh process.

That's not a mistake I will make again, so this should be a non-issue.

By "stuck", I mean a process which either has no children and is just sitting
around doing nothing or a process which is running something like cvs but has
been running for 24 hours.  I believe that most of these are due to the people
running behind a firewall.  I'd hoped that this problem would go away with a
newer version of linux and newer version of ssh but I guess it won't
just magically disappear.  So, I'll be investigating some
/proc/sys/net/ipv4 settings this weekend.

cgf

Re: ssh crash explained

[Frank Ch. Eigler]

Hi -


cgf wrote:

> [...]
> By "stuck", I mean a process which either has no children and is just sitting
> around doing nothing or a process which is running something like cvs but has
> been running for 24 hours.  I believe that most of these are due to the people
> running behind a firewall.  [...]
> So, I'll be investigating some /proc/sys/net/ipv4 settings this weekend.

If you suspect connection glitches, consider adding to /etc/ssh/sshd_config
settings like ...

ClientAliveInterval 120
ClientAliveCountMax 5
KeepAlive yes # we already have this one


- FChE