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
{{ message }}
This repository has been archived by the owner on Aug 13, 2019. It is now read-only.
updating the xsconsole RPM restarts the xsconsole service.
Thus it kills any child process of the console, such as an interactive shell opened from there. Such shell could be running an important task, which gets killed.
In XCP-ng, we found out about this during the development phase of XCP-ng 8.0.0, because we updated the xsconsole RPM: right after installing the host, it is recommended to run yum update to get the latest updates, so one might open a shell from the console, run yum update, and when it reaches the point where the %postun scriptlet of the xsconsole RPM is executed, XSConsole restarts and the yum command is killed by that. This leaves the RPM database in an inconsistent state because the old xsconsole RPM cannot be removed (nor some other upgraded RPMs that were waiting for their turn to remove the old versions that had been upgraded). Then you need to install yum-utils and run yum-complete-transaction.
I think the same issue could occur if someone would open a shell from XSConsole to install a Citrix Hypervisor update from the update ISO if that update were to contain an updated xsconsole RPM.
I'm not sure about the best fix. Maybe avoid restarting the service and drop a file in /tmp instead to let a running XSConsole know that it should restart when it gets the opportunity to do it without killing subprocesses or running tasks?
The text was updated successfully, but these errors were encountered:
Due to
xsconsole/mk/xsconsole.spec.in
Line 37 in bf2ce68
Thus it kills any child process of the console, such as an interactive shell opened from there. Such shell could be running an important task, which gets killed.
In XCP-ng, we found out about this during the development phase of XCP-ng 8.0.0, because we updated the xsconsole RPM: right after installing the host, it is recommended to run
yum update
to get the latest updates, so one might open a shell from the console, runyum update
, and when it reaches the point where the %postun scriptlet of the xsconsole RPM is executed, XSConsole restarts and the yum command is killed by that. This leaves the RPM database in an inconsistent state because the old xsconsole RPM cannot be removed (nor some other upgraded RPMs that were waiting for their turn to remove the old versions that had been upgraded). Then you need to installyum-utils
and runyum-complete-transaction
.I think the same issue could occur if someone would open a shell from XSConsole to install a Citrix Hypervisor update from the update ISO if that update were to contain an updated xsconsole RPM.
I'm not sure about the best fix. Maybe avoid restarting the service and drop a file in /tmp instead to let a running XSConsole know that it should restart when it gets the opportunity to do it without killing subprocesses or running tasks?
The text was updated successfully, but these errors were encountered: