GLPI Agent 1.13 Deploy Task Redeployment

[CSpencerLTD]

Bug reporting acknowledgment

Yes, I read it

Professional support

Yes, I know

Describe the bug

We have ran into an issue with GLPI Agent 1.13 where deploy tasks seem to run, but then re-run when the agent next checks in again (they keep re-running when the agent checks in).

Has anyone else seen this issue? We are unsure if its a server sided issue or an agent issue, but GLPI Agent 1.11 did not have this issue.

To reproduce

If there are any tasks deployed to inventory groups, force inventory on the agent, wait a couple of minutes and refresh the web interface on GLPI (Computer > Tasks), you should see a list of prepared tasks which have previously deployed.

Expected behavior

The tasks should only deploy once.

Operating system

Windows

GLPI Agent version

v1.13

GLPI version

10.0.16

GLPIInventory plugin or other plugin version

GLPI Inventory v1.5.0

Additional context

No response

[g-bougard]

Hi @CSpencerLTD

can you share the difference you can see in glpi-agent log between the v1.11 and v1.13 runs and which gives you the confidence something no more works ?

[CSpencerLTD]

Please see attached log files. 1.13 seems to have a lot of errors.

glpi-agent_1.11.log

glpi-agent_1.13.log

I have removed our domain name from the log files.

[g-bougard]

Hi @CSpencerLTD

errors like [error] [http server] unknown path: are not glpi-agent problems. These lines are only a trace a someone trying to abuse the agent on unsupported path on glpi-agent HTTPd UI interface. There's probably a process somewhere which is trying to scan the agent for webserver security hole. GLPI-Agent just reports it can't handle that requests. Have you this agent exposed on internet ? If yes, be sure to have a secured definition in httpd-trust configuration. If no, this may be due to an internal process trying to detect any wellknown security hole to alert network team.
Anyway, you can ignore that message regarding your question.

I see something you don't have in v1.11:

[Wed Apr  2 08:10:43 2025][info] usercheck before: current job continued by user choice
[Wed Apr  2 08:11:01 2025][info] usercheck before: current job continued by user choice
[Wed Apr  2 08:11:24 2025][info] usercheck after: current job continued by user choice
[Wed Apr  2 08:11:56 2025][info] usercheck before: current job continued by user choice
[Wed Apr  2 08:14:24 2025][info] usercheck after: current job continued by user choice

This tells me the deployed package is defined with user interaction. Depending on the definition, this can involve the package needs to be deployed again later. Maybe this explains what you think is an issue.
I didn't see that case in your v1.11 log. Are you using the same package in the 2 logs ?

[CSpencerLTD]

Hi @g-bougard,

Yes, we have some packages which require the application (we use packages for application updates mainly) to be closed first.
So, we have configured some packages to first display to the user a warning that they have an hour to close the affected application down, or after 1 hour, GLPI agent will terminate the relevant application and then proceed with the updates - which works perfectly.

Since we have pushed out GLPI Agent 1.13 even packages that do not have user interactions configured have gone in to a 'prepared' state despite the package already completing successfully on the user's pc.

From what I could see, it looked as if the automatic action 'taskscheduler' had caused the packages to be re-deployed but I am confused as to why, as this was not an issue on earlier versions of the agent/server. As a quick fix, I updated all of the tasks on GLPI Inventory to Active=No just to prevent any other packages from deploying again.

Our GLPI database is very old (from early versions of GLPI) so we have installed it onto a new server, new database etc, everything is up to date. We will test a few PC's on the new GLPI Server and see if the same bug remains (we suspect its something to do with our current Server)

Kind Regards,

Connor

[g-bougard]

Hi @CSpencerLTD

okay, thank you for the clarification.

I'll move this issue as a discussion if you want to update the history. Actually, I don't see any evidence of an issue on glpi-agent itself.