[SOLVED] EVE-NG pro cluster : node shutdown issue

[mamakos]

Hello,

EDIT : this is solved since version 4.0.1-86 (the 1st of may 2022)

Here is our cluster setup :

• Bare Metal installation

• EVE Pro 4.0.1-65

• 1 master, 4 satellites (3 are really in use)

I conducted the tests in a situation where the different cluster members were not overcrowded, I swear I did I give the figures below

We have nodes (Nokia) that share the same UUID for their license is attached to one specific UUID. So nodes are using the same UUID in order to be able to use the same license. This is working fine. Nodes start properly and can be used without any major issue except when we try to stop one node :

• On master we can stop each node independently without any issue.

• On satellite member, when stopping one node, all the nodes that share the same UUID get stopped in all labs of all users. This is quite annoying although we can play trick to our colleagues

I moved the nodes to one satellite to another, this makes no difference.

In the /opt/unetlab/data/Logs/php_errors.txt file we can see that the unl_wrapper CLI is properly specifying the device ID (14 for instance here)
[28-Jan-2022 13:37:35 Europe/Berlin] stop node with sudo /opt/unetlab/wrappers/unl_wrapper -a stop -m 3 -T 1 -D 14 -F "/opt/unetlab/labs/Users/ICC/LOUIS/test.unl" 2>> /opt/unetlab/data/Logs/unl_wrapper.txt

For the record, the hosts/vCPU/RAM distributionbefore I started 8 nodes.

Code: Select all

HOSTS distribution
user  |  cluster  | muztc628  | muztc629  | muztc620  |
total |  11 hosts |   3 hosts |   8 hosts |
  2   |  10 hosts |   2 hosts |   8 hosts |
  1   |   1 hosts |   1 hosts |           |


vCPU distribution
user  |    cluster    | muztc628/112  | muztc629/112  |  muztc620/96  |
total |       19 vCPU |        5 vCPU |       14 vCPU |
  2   |       18 vCPU |        4 vCPU |       14 vCPU |
  1   |        1 vCPU |        1 vCPU |               |


RAM distribution
user  |   cluster       | muztc628/385574 | muztc629/385574 | muztc620/257395 |
total |        43008 Mb |        10240 Mb |        32768 Mb |
  2   |        40960 Mb |         8192 Mb |        32768 Mb |
  1   |         2048 Mb |         2048 Mb |                 |

I can share the logs of starting/stopping nodes on master or satellite if you want.

Is there a solution to avoid such a behavior (except using different UUID of course).

Thank you !

[ecze]

Thanks for reporting.

This issue will be resolved in the future.
As uuid should be unique normally, this usecase is a bit special.
However, the code will be modified to permit such.
Could you send us log from sats ? ( journalctl -u 'eveng*' ) and share uuid to verify in log...

Kind Regards,

E.

[mamakos]

Great. I won't be slashed by my colleagues anymore then
Thank you ! We look forward to upgrading to the next release


Edit : May the 6th
I realized I had forgotten to send you the logs you requested
So I sent them via PM

[mamakos]

Marked the topic as solved.
Version 4.0.1-86 solved the issue