Seems that there are issues with sessions' start. And from Terminal commands and from PY scripts.
"File "/usr/lib64/python2.7/site-packages/xpra/net/protocol.py", line 993, in steal_connection assert not self._closed, "cannot steal a closed connection" AssertionError: cannot steal a closed connection"
Xpra: Fatal IO error 11 (Resource temporarily unavailable) on X server :4. (gedit:3970): Gdk-WARNING **: gedit: Fatal IO error 11 (Resource temporarily unavailable) on X server :4.
As always, please follow wiki/ReportingBugs. In particular:
Centos 7.x The easiest way is to put a call for a new session for user on target hosts in xx_auth.py/get_sessions() on Proxy server. Basically 1 connection can be established normally but then there are more than 1 session on 1 target_host - behavior of proxy/sessions/ports is unpredictable (mentioned above).
Version 2.0.2 reliably starts many sessions with ports open less then 3 sec for users. Proxy connects html client easily. With the same ssh paramico call from the same xxx_auth.get_session() function. Tested today and 2.1 and 2.0 .2 on the same server and 2 auth modules(multi and sql). Seems that it could be somehow linked with new in 2.1 "flag of stealing session" or other modif.
P.s. almost happy :-) as it got me absolutly mad :-) so getting back to normalize balancing script
Please provide reproducible steps for this issue with any supported xpra version, without making changes to the code. Otherwise I will have to close this ticket as "invalid". I am also having a very hard time understanding your comments.
Having tried with different scenarios I was unable to reproduce that with Terminal use only. So the all descriptions are relevant only to the calls from the modified code.
this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/1506