You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Upgraded Nomad cluster from 0.12.5 -> 1.0.4. After the upgrade experienced a behavior change in the users running with exec driver. In nomad version 0.12.5, each nomad task received a unique PID for each worker process. In Nomad version 1.0.4, each nomad task has PID 1. This is a breaking change in our scenario because we are using Nomad with a 3rd party application to run multiple instances of the application on the server. The application records to a database the server name and PID and it expects the servername/PID to be unique and when it is not unique, the application crashes. We do not have access to change the 3rd party code that is crashing. Is there some setting that I can set to revert Nomad 1.0.4 to assign a unique process id to each task?
As an alternative solution, we could change the driver type from exec -> raw_exec but that would require quite a bit of work because we are using the memory constraints, volume mapping, and chroot functionality of the exec driver.
The text was updated successfully, but these errors were encountered:
Hi @LukeMauldin! This behavior was changed in #9911 which shipped in Nomad 1.0.3 to prevent CVE-2021-3283. But we added a configuration flag in #9982 and #9990 to opt-in to restoring the previous behavior, which we shipped in 1.0.4.
So you can restore the previous behavior by setting default_pid_mode = "host" in the exec driver plugin configuration on your agents.
I'm going to lock this issue because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.
Nomad version
Output from
nomad version
:1.0.4
Operating system and Environment details
Linux x64 - RHEL 8.3
Issue
Upgraded Nomad cluster from 0.12.5 -> 1.0.4. After the upgrade experienced a behavior change in the users running with
exec
driver. In nomad version 0.12.5, each nomad task received a unique PID for each worker process. In Nomad version 1.0.4, each nomad task has PID1
. This is a breaking change in our scenario because we are using Nomad with a 3rd party application to run multiple instances of the application on the server. The application records to a database the server name and PID and it expects the servername/PID to be unique and when it is not unique, the application crashes. We do not have access to change the 3rd party code that is crashing. Is there some setting that I can set to revert Nomad 1.0.4 to assign a unique process id to each task?As an alternative solution, we could change the driver type from
exec
->raw_exec
but that would require quite a bit of work because we are using the memory constraints, volume mapping, and chroot functionality of theexec
driver.The text was updated successfully, but these errors were encountered: