FOG 1.6.0.158 - Multicast error (session has already started with 0 clients)
-
Hi everyone,
I’m running FOG 1.6.0.158 in a Hyper-V environment on an isolated (internal) network. My DHCP server is Windows Server 2022, and the boot file is ipxe.efi. I’ve tried this on two different Hyper-V hosts; on the second host I also installed the latest dev-branch version, and the issue occurs on both 1.6.0.158 and the dev-branch.
The problem: I create a multicast session with 2 clients and a 10-minute timeout. Then I PXE boot the machines and try to join the multicast session, but I get the message: “That session has already started and can no longer joined.” Meanwhile, the FOG web UI shows 0 connected clients under Tasks / Active Multicast Tasks.
I installed FOG using the default standard install mode (–install-mode standard), and at the end of the installation I let the installer open the required ports in firewalld. The OS is AlmaLinux 9.8, and SELinux is set to permissive. Both servers are fresh installs on a clean AlmaLinux 9.8 system, not upgrades from 1.5. I’ve attached 2 screenshots and a few lines from the multicast log.
Any help solving this would be greatly appreciated.
Thanks,
Miklós

P.S. If you need any other logs, or the status of the services or the multicast session, just let me know and I’ll post them in a reply.
-
@Mikeee89 Thanks for the log, it shows the cause. The first line of the udp-sender command has
--max-wait 10. That value is seconds, not minutes.The Session Timeout field on the Multicast page asks for minutes, but FOG stored the number as seconds. So your 10-minute session waited only 10 seconds after it started. After that, the PXE menu treats the session as already started and refuses the join, even with 0 clients connected.
The fix is in https://github.com/FOGProject/fogproject/pull/2114 and goes to dev-branch.
Until you update, either of these works on your current install:
- Enter the timeout in seconds. For 10 minutes, enter 600.
- Or leave the timeout blank. FOG then uses the global setting FOG_UDPCAST_MAXWAIT (in minutes) and converts it correctly.
Cancel the stuck session first, then create a new one.
-
PS: i used this dev-branch version: 1.6.0.208
-
@Mikeee89 Thanks for the log, it shows the cause. The first line of the udp-sender command has
--max-wait 10. That value is seconds, not minutes.The Session Timeout field on the Multicast page asks for minutes, but FOG stored the number as seconds. So your 10-minute session waited only 10 seconds after it started. After that, the PXE menu treats the session as already started and refuses the join, even with 0 clients connected.
The fix is in https://github.com/FOGProject/fogproject/pull/2114 and goes to dev-branch.
Until you update, either of these works on your current install:
- Enter the timeout in seconds. For 10 minutes, enter 600.
- Or leave the timeout blank. FOG then uses the global setting FOG_UDPCAST_MAXWAIT (in minutes) and converts it correctly.
Cancel the stuck session first, then create a new one.
-
The issue affects both FOG 1.5.10.2534 and FOG 1.6. Version 1.5.10.2518 is not affected. The error no longer occurs when entered as seconds. Thanks for the help.
-
M Mikeee89 has marked this topic as solved
-
@Mikeee89 Good to hear that seconds work. The fix is now merged into dev-branch, so a dev-branch update makes the field take minutes again.
1.5.10.2518 and 1.5.10.2534 have the same multicast code, so this bug cannot be in one and not the other. 1.5 has no per-session timeout. Its session waits FOG_UDPCAST_MAXWAIT minutes (10 by default).
Can you post the
udp-senderline from /opt/fog/log/multicast.log on the 1.5.10.2534 server? The--max-waitvalue on the first stream shows what that server waited. -
@Tom-Elliott I checked again. I was mistaken; FOG 1.5 was not affected. Thanks again for the help.