if ioc reboots, then WeTest should retry last stage test if the reason of the failure was "unable to connect to setter..."

[vnadot]

Description of the case:

• Every test stage was a success up to stage "400". Then the ioc rebooted during stage "400". So WeTest raises an alarm on this test stage "unable to connect to setter...". Then the ioc restart and connection is regained. Finally WeTest keeps going on other test stages (800.0 and co) and it works fine.

What I'm expecting:

• if communication is regained, retry stage "400" if the reason of the failure was "unable to connect to setter..."

[gohierf]

Hi,

How can WeTest know that the issue is that the IOC has rebooted ?

WeTest only interfaces with the PVs using caget and caput. I'm not sure how it could detect whether the IOC is restarting, or there is a network issue, or an issue with the PV name.

Aside from this specific situation, we could implement a feature that goes like this:

• test as failed (using retry or not)
• test is paused (using on_failure with pause to have user input)
• the user can currently select to continue or abort, we could add the option to retry which is the same as continue but we run the last test again.

[gohierf]

I fail to see how this issue is tagged as "bug", WeTest did try to set and get a PV which didn't aswer, that is the expected behavior isn't it ?

[vnadot]

I think that with pyepics you have some callback about the PV connection. I think that Stéphane is using that for gengiscan. If you are doing a test and get at the same time a connection issue (PV was connected before), you can just pause/retry the test.

I put the "bug" label because a PV connection issue is, to me, not a reason to make the test failed.

[stephane-cea]

Hi,

Actually with pyepics you might want to simply use the timeout parameter of the get(...) function, like so:

import epics

pv = epics.pv.get_pv("pv_name")
value = pv.get(timeout=timeout)

The timeout will indicate the maximum time to wait (in seconds I think) for data before returning None (more details here).

With a long enough timeout, would it be OK?