Mender-monitor reports disconnected after OTA

General Discussions
1

I have a user who reports that occasionally, after doing an OTA update, mender-monitor stops reporting alerts. It looks like, in the failing case, the mender-monitor logs report being disconnected even though the device has in fact connected. A restart of the mender-monitor daemon gets everything working again.

Are there any known issues or workarounds? I’m trying to get a more reliably reproducible scenario but at the moment this is all I have.

Drew

2

@kacf @merlin any thoughts here?

3

apologies for the delay @drewmoseley we have been at the company summit for the last week.

I will get you @merlin 's attention here.

And, this is not a bug that I’m aware of at least.

4

thank you, and sorry, I am a bit late with the answers to the hub (got back at 2am yesterday, started at 6am, sleeping in mid-clicks).
thanks Drew for the report, do we know what kind of update was that? if you could give me some details, I will try to reproduce. At least and if you could: if it was the full rootfs update, and what kind of checks were enabled with the Monitor add-on.
if not I will try to enable one log check with streaming logs and one with a file, and do a full rootfs update.
is it possible that the update the customer is performing overwrites the file that is being monitored?

peter

5

Hi guys,

Hope the summit was good. It’s definitely nice to be getting out in person again.

I don’t have much specific to report on this. It does not happen every time and we’ve not been able to narrow it down beyond something that just occasionally happens. The user is only doing full image updates so nothing in the active filesystem is getting updated.

The checks enabled are a few service scripts and a few log triggers.

For now we have just implemented a workaround of restarting mender-monitor and that seems to work but obviously we want to understand it a bit more. If we come up with any more details, I’ll share them here.

Drew

6

thanks. it was.
thank you. if I am able to reproduce the behaviour, I will update this thread.
if you happen to have the client and mender-monitor logs, that would help. (you can share them privately if you prefer)

pg