Message boards :
Theory Application :
Theory native tasks fail when restarted
Message board moderation
| Author | Message |
|---|---|
 Send message Joined: 21 Feb 11 Posts: 83 Credit: 577,613 RAC: 0  
|
They fail with msg="runc run failed: container with given ID already exists" https://lhcathomedev.cern.ch/lhcathome-dev/result.php?resultid=3338978 |
 Send message Joined: 15 Jun 08 Posts: 2632 Credit: 271,713,117 RAC: 93,186  |
Describe the steps you have done to stop/restart the task. Is your BOINC client configured to "leave non GPU applications in memory"? |
|
Send message Joined: 21 Feb 11 Posts: 83 Credit: 577,613 RAC: 0
|
sudo systemctl stop boinc-client sudo systemctl start boinc-client or sudo systemctl reboot |
|
Send message Joined: 15 Jun 08 Posts: 2632 Credit: 271,713,117 RAC: 93,186 |
Theory native tasks do not survive a BOINC restart or a reboot. |
|
Send message Joined: 21 Feb 11 Posts: 83 Credit: 577,613 RAC: 0
|
Before cgroupsv2 they restarted from beginning. Now they just crash. |
|
Send message Joined: 15 Jun 08 Posts: 2632 Credit: 271,713,117 RAC: 93,186 |
Native apps (ATLAS/Theory) were never intended to fully replace vbox apps. They were developed to run efficiently on datacenter hosts (24/7) with rare planned shutdown breaks. The Theory cgroups v2 version adds a functionality to pause/resume without the need to manually fiddle around with cgroups. This makes it a bit more cooperative on hosts running Non-Theory tasks but it never claimed to solve all basic issues from above. As a user you have to decide what is more important for your system: - run it 24/7 and plan shutdowns (hence stop running new native Tasks early enough) - regularly stop/restart running tasks (then run vbox apps) |
©2025 CERN