[Parti-discuss] reconnecting to an abandoned Xvfb-for-Xpra server

[Thomas Käfer]

The servers home directory (or better said the / directory) is on
/dev/md0 (a mdadm raid1 device).

Thanks for the fix. I'm gonna try it the next time I have a dead
session, I already killed them this time and started over.

Regards, Thomas

2009/5/21 Nathaniel Smith <njs at pobox.com>:
> 2009/5/20 Thomas Käfer <kaefert at gmail.com>:
>> I found some message of parti-discuss where you discussed the problem
>> of dead sessions, and how to restore them. I tried to follow the
>> procedure, but without success. See xpras output in the attached txt
>> file.
>
> Yes, uh, 'xpra upgrade' is broken in 0.0.6. *hides head*
>
> You can hack it by opening up your copy of xpra/server.py (NOT
> xpra/scripts/server.py) and finding lines 234-241, which should look
> like:
>
>        ### Load in existing windows:
>        for window in self._wm.get_property("windows"):
>            self._add_new_window(window)
>
>        for window in get_children(root):
>            if (is_override_redirect(window) and is_mapped(window)):
>                self._add_new_or_window(window)
>
> and moving this block of code down to the bottom of the function, just after
>
>        ### All right, we're ready to accept customers:
>        self._protocol = None
>        self._potential_protocols = []
>        for sock in sockets:
>            self.add_listen_socket(sock)
>
> (Make sure to preserve indentation.)
>
> It's also fixed in current mainline.
>
>> If someone has a hint for me, what I did wrong, or what I could
>> do to prevent xpra sessions from dying, I would be grateful.
>
> I'm quite confused about your problem with sessions dying. From the
> log file you sent me, it looks like they aren't dead at all -- but yet
> 'xpra list' thinks that they are. I've seen something like that when
> the home directory was on NFS and I accidentally ran 'xpra list' on
> one machine while the actual server was running on another... I don't
> suppose that you're doing something like that?
>
> -- Nathaniel
>