You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In one particular situation there is a collabnet.service that wants to run more "systemctl start" routines on other services. Upon a container start, both the collabnet.service and the other services are run by PID 1 as they are all enabled. As a result there may be two processes of the same service being run parallel to startup.
Now this situation has some obvious problems attached to it. Not even about possibly destroying data, but I noticed it first by having PIDFile knowing only the last process that has started, so a shutdown of the processes will not be complete when the container needs to shut down. A "systemctl halt" will leave some processes running. And that's a general problem with other circumenstances where a "start" call is issused in parallel from different sides.
The text was updated successfully, but these errors were encountered:
Using lock files for each service does only solve the problem partially - one has to check that an active service is not followed by another start sequence, and that can also happen if a service is run as type=forking where it takes some time till the actual program is up.
In the real world it did help to make the collabnet.service start to be much more stable.
In one particular situation there is a collabnet.service that wants to run more "systemctl start" routines on other services. Upon a container start, both the collabnet.service and the other services are run by PID 1 as they are all enabled. As a result there may be two processes of the same service being run parallel to startup.
Now this situation has some obvious problems attached to it. Not even about possibly destroying data, but I noticed it first by having PIDFile knowing only the last process that has started, so a shutdown of the processes will not be complete when the container needs to shut down. A "systemctl halt" will leave some processes running. And that's a general problem with other circumenstances where a "start" call is issused in parallel from different sides.
The text was updated successfully, but these errors were encountered: