#172 closed enhancement (fixed)
ability to upgrade xpra server sessions in place
Reported by: | pmarek | Owned by: | Antoine Martin |
---|---|---|---|
Priority: | minor | Milestone: | 0.12 |
Component: | Server | Keywords: | |
Cc: |
Description (last modified by )
At the moment, if one uses "xpra upgrade
" on an existing xpra session, winswitch will see that the server process it had started has now terminated and it will close the session and kill all the remaining session processes.
Options:
- detect an exit code from the xpra server which will mean "upgrade me" and re-spawn it.
This could be tricky as there are daemons involved, pipes and file descriptors, gio file watchers, etc.
- change the xpra code to exec the new process in place so we keep the same pid and this becomes transparent to winswitch
Another benefit is that all the start options would then be preserved too (password, etc) - and we could even envisage that extra options could be specified to override existing ones
Attachments (1)
Change History (7)
comment:1 Changed 13 years ago by
Milestone: | → 1.1 |
---|---|
Owner: | changed from Antoine Martin to Antoine Martin |
Status: | new → accepted |
comment:2 Changed 13 years ago by
Resolution: | → worksforme |
---|---|
Status: | accepted → closed |
comment:3 Changed 13 years ago by
Description: | modified (diff) |
---|---|
Milestone: | 1.1 → 0.12 |
Resolution: | worksforme |
Status: | closed → reopened |
Summary: | "winswitch upgrade" → ability to upgrade xpra server sessions in place |
comment:4 Changed 13 years ago by
Description: | modified (diff) |
---|
comment:6 Changed 13 years ago by
Yesterday I got my xpra killed on upgrade.
Here's part of the winswitch server logfile, perhaps there's something in it.
Note: See
TracTickets for help on using
tickets.
edit: the original bug made it sound like the problem was with upgrades to winswitch, but in fact the problem is with xpra sessions exiting on upgrade.
Restarting the winswitch server should not lose any sessions and if your client is set to auto-resume, it should re-connect the sessions too in the event that some of them drop (which they should not do - as each session uses its own daemon)