[squid-users] Squid.out reports errors for lines that do not exist in squid.conf

Amos Jeffries squid3 at treenet.co.nz
Thu Jun 7 10:11:51 UTC 2018


On 07/06/18 18:13, Eliezer Croitoru wrote:
> Amos,
> 
> Systemd can be define to run a specific command for a "reload" and even if nobody wrote the line in a service file it's there since almost day one of systemd services.
> 

*If* that mechanism is used there is no difference in the commands. If
it is not used, the systemd ones are actively dangerous. So no harm in
advising the safe one be used in either case.


> And.. if the version is el6 I believe it's still a sysVinit based system.

Cheadle was using systemd's "service ..." commands. Which I am advising
to avoid because something indeterminate is going wrong with the config
loading and startup process. If the OS is actually SysV those systemd
commands are even more inappropriate.


> Squid -kparse should detect and squid -kreconf should resolve any issue if it's not a fatal one that stopped the service.
> 
> Not directly related but.. only if Squid doesn't release at all any memory it catches then a restart would be a must at some point.
> From what I have seen in the 2.7 and 3.x code in the past it seems that there should be some level of memory cleanup\release.
> Also I have systems that has up-time of almost a year so I am a bit confused why should a restart would be requied?

The admin has apparently got themselves into a difficult situation and
it is no longer clear whether systemd or Squid master process is in
control of the worker processes which are running and with what config.
They both fight over "service ..." commands.

The only thing which is guaranteed to restore Squid to known state is a
full shutdown. Ensuring that everything has exited before proceeding
with the start action.

Amos


More information about the squid-users mailing list