Hi guys,
Just created a new RDS host with 8.5.0 agent (--multiple-session) and corresponding server farm and desktop pool in Horizon console. The host is Ubuntu 20.04.
However all connection tests fail, and this error is logged at /var/log/vmware/viewagent-debug.log
2022-05-04T16:20:22.693Z ERROR <pool-3-thread-4> [DesktopManager] Failed to start session for user john doe, the desktop is in start storm status.
So, where I should begin my troubleshooting?
Cheers,
Perttu
Seems GDM failed to start with XDMCP.
You can restart GDM by
sudo systemctl restart gdm3
And then check if XDMCP enabled by GDM by
sudo lsof -i:xdmcp
If it still doesn't work please try to reboot and see if GDM can be back to work with XDMCP.
Hi Perttu
Could you descript your test case? Does all the connect session failed ? "the desktop is in start storm" means your start too many connection sessions at a short time, so the desktop agent refuse continue connections. If there isn't any connection successful, I think it is a bug for multiple sessions, and could you update the detail Linux Desktop Agent log? If there are some connection successful, but some connection successful, I think you can wait a moment and retry the connection again.
Best Regards
Kai
Hi,
Test is just a simple login from Horizon Client. There are no other users. The viewagent-debug.log includes real hostnames and IP addresses so I wish to not share it publicly here without throughout sanitise. I can send it privately if needed. But here is the a one complete login snippet:
2022-05-05T14:38:40.388Z INFO <main> [StandaloneAgent] Agent started done
2022-05-05T14:38:42.535Z DEBUG <SwiftMQ-SessionPool-3> [msgid] Validating message with ID: 'null'.
2022-05-05T14:38:42.546Z DEBUG <pool-3-thread-1> [StandaloneAgent] rds-ubuntu received: StartSession[id:-3cff4649:1802a4ee57a:5289|userDn:CN=S-1-5-21-<sanitized>,CN=ForeignSecurityPrincipals,DC=vdi,DC=vmware,DC=int|userName:johndoe|domainName:null|sessionType:DESKTOP|connectionId:b8d3e56f_98c2_43db_9a1a_e00a1520ce52|locationId:<sanitized>|brokeringEndpointId:null|securityGatewayId:uag.example.com|ndcContext:SESSION:b8d3_***_ce52|protocol:BLAST|externalPort:-1|appId:null|appLbPattern:null|appLbCount:0|appExec:null|appStartDir:null|appArgs:null|filePaths:null|appMaxed:null|sessionTtl:42769955|requestTtl:-1|userGuid:null|certSsoData:null|isPreLaunch:false|isUnauthenticatedAccess:false|enableCollaboration:false]
2022-05-05T14:38:42.546Z DEBUG <pool-3-thread-1> [DesktopManager] Current total desktops count: 2, pending desktops count: 0, initing desktops count: 2, ready desktops count: 0, logging desktops count: 0, logged desktops count: 0, unlogging desktops count: 0
2022-05-05T14:38:42.546Z DEBUG <pool-3-thread-1> [DesktopManager] [DesktopID: 3] Pending desktop to avoid start storm.
2022-05-05T14:38:42.546Z ERROR <pool-3-thread-1> [DesktopManager] Failed to start session for user johndoe, the desktop is in start storm status.
2022-05-05T14:38:42.547Z DEBUG <pool-3-thread-1> [ComponentResponse] Response directed to:ID:/127.0.0.1/-2217732850099084150/606339/0
2022-05-05T14:38:42.547Z DEBUG <pool-3-thread-1> [ComponentResponse] Message is <?xml version="1.0"?><TERMINALRESPONSE> <ID>-3cff4649:1802a4ee57a:5289</ID> <SERVERDN>cn=59312694-12dd-4c7f-90c5-48a524c64440,ou=servers,dc=vdi,dc=vmware,dc=int</SERVERDN> <SERVERPOOLDN>cn=rds-ubuntu,ou=server groups,dc=vdi,dc=vmware,dc=int</SERVERPOOLDN> <SERVERDNSNAME>rds-ubuntu
</SERVERDNSNAME> <ERROR-CODE>AGENT_ERR_MAXSESSIONS</ERROR-CODE></TERMINALRESPONSE>
2022-05-05T14:38:42.548Z DEBUG <pool-3-thread-3> [TopicMessageResponder] Response sent.
2022-05-05T14:38:59.801Z INFO <Service Shared Timer> [StandaloneAgent] -----SVI State: false USE SVI: false
2022-05-05T14:38:59.801Z DEBUG <Service Shared Timer> [ComponentResponse] Message is AGENT_NOTIFICATION
2022-05-05T14:38:59.803Z DEBUG <pool-3-thread-2> [AbstractTopicPublishingManager] Message published.
2022-05-05T14:39:32.734Z INFO <Thread-9> [DesktopDaemonIPCHandler] DesktopDaemonIPCHandler get IPC message
2022-05-05T14:39:32.737Z DEBUG <Thread-9> [DesktopManager] [DesktopID: 1] Handle IPC event
2022-05-05T14:39:32.740Z DEBUG <pool-4-thread-1> [DesktopHandler] [DesktopID: 1] Submit event task
2022-05-05T14:39:32.743Z DEBUG <pool-4-thread-1> [DesktopHandler] [DesktopID: 1] Process desktop daemon ipc event: DesktopDestroyed
2022-05-05T14:39:32.744Z DEBUG <pool-4-thread-1> [DesktopHandler] [DesktopID: 1] Desktop was destroyed now.
2022-05-05T14:39:32.745Z INFO <pool-4-thread-1> [DesktopHandler] [DesktopID: 1] Retry to prepare desktop 1 times.
Hi,
The test was just a single login from Horizon Client. There are no other users/login at the same time. The desktop doesn't work for any login. I tried to attach a log snippet here but the post was marked as spam, so if full log is needed here I have to send it privately.
Also in /var/log/Xorg.100.log is logged:
[ 312.217] (II) VMWMEM(0): Setting screen physical size to 304 x 228
[ 438.297] XDM: too many retransmissions
[ 438.299] (II) Server terminated successfully (0). Closing log file.
Regards,
Perttu
Hi,
Few more details, in /tmp/vmware-root/vmware-DesktopWorker-9559.log is logged
2022-05-05T12:14:02.593Z In(05) host-9576 [Child Process]: display manager invalid, XDMCP port isn't enabled
but in /etc/gdm3/custom.conf as normally:
[xdmcp]
MaxSessions=50
DisplaysPerHost=50
Enable=true
I don't know whether these are related.
Seems GDM failed to start with XDMCP.
You can restart GDM by
sudo systemctl restart gdm3
And then check if XDMCP enabled by GDM by
sudo lsof -i:xdmcp
If it still doesn't work please try to reboot and see if GDM can be back to work with XDMCP.
Thank you,
I had restarted the machine several times already, but this hint lead me to right direction. I clearly have some issues with gdm3 autostart on startup.
Umm, should gdm3 be a dependency for viewagent service, so it would be assured, that gdm3 is running before viewagent starts up.
My pleasure.
Yes, regarding service dependencies, we had considered it. We implemented a weak dependency by specifying `After=display-manager.service' in viewagent.service.
In addition, viewagent startup script also performs a timeout check on the display manager and gives error log in case of failure. You can find it in /var/log/vmware/viewagent-debug.log by keyword 'display manager'.
The reason we do not specify the stronger one `Requires=` is to avoid the potential affect of existing sessions since viewagent will be restarted(stopped) if GDM explicitly stopped (or restarted).