BUG: Cannot break import of plate

[pwalczysko]

In Insight, import data_repo/public/images/HCS/INCELL2000 - select first subfolder there and import.
When trying to cancel or cancel all during the scanning phase (fairly early after import started, ~2 mins, the cancelling action is being shown as proceeding in Importer in the line, but the spinner is still spinning (both in importer and Data Manager). See screenshots.
Importer can be closed, but the Insight as a whole cannot be Quitted (Force Quit necessary).

[jamoore]

Before I forget, one of the major reasons for having this properly, was that Jason suggests an important metric (if not *the* metric) is: how much faster is FS to allow me to close my client for a particular import. I.e. if the server is off working, I don't have to care.

[pwalczysko]

Noted another phase where the Cancel process is not smooth: the test_images_good/lei/050118.lei has on FS a long-ish "preparing upload" phase, during which the Cancel button is not greyed out, but if you press it , nothing happens.

[jamoore]

Starting to look at this now that the inplace import API & hooks are in place.

[jamoore]

HCS import under linux using dev_5_0 (~5.0.0.rc1)

[jamoore]

Uploaded screenshot which shows a grayed "cancel" button on upload. This may not be the correct location for the information, but I wanted to capture it while testing this issue.

Looking at the rest of Petr's feedback, it looks 1) these are largely state-issues (when can X be pressed) and therefore 2) probably make testing the feature I'd like to target for 5.0.0 from comment:1 not testable in insight. If there are no objections, I'd push this ticket to 5.0.1 or similar, and then test a cancel workflow from the CLI to see if the background importing continues as advertised.

[jamoore]

Handling the CLI info under ​https://trac.openmicroscopy.org.uk/ome/ticket/11932#comment:2 and pushing this for UX/GUI investigation.

[jburel]

Issues noticed while testing the CLI in place import during the road trip. Cancelling scanning is something we have to address for such files. (cf. North America report)
Nothing we can do for 5.0.1, Pushing to 5.0.2, so we start the discussion asap.

[jburel]

Pushing, nothing we can do for 5.0.2 since we cannot cancel during the scanning phase.

[jburel]

ticket linked to ​https://trello.com/c/kTT3vVys/48-import-tng

[jamoore]

Anything to be done here for 5.1.1?

[pwalczysko]

The behaviour did not change here as checked today with trout merge.

[jburel]

This is not only limited to "plate"

"scanning" cannot be broken, i.e. BF tries to find out the files composing the image. Option:

• break Importcandidates
• Upload w/o scanning and do the scanning server side.

No realistic to have that implemented for 5.1.1

[jburel]

Discussed with Josh today, moving to 5.1.2

[jamoore]

Melissa: specifically with the thought that we could make some progress on the callback/scanning side of things in BF.

[jamoore]

I'll look into what such an API could look like with an eye on keeping it as non-invasive as possible.

[jamoore]

Started work on the API, but obviously nothing for 5.1.2. Pushing.

[jamoore]

Requires new API; pushing.

[jamoore]

​https://github.com/bitronix/btm/pull/46/files#diff-5997179544172346b7113a96958610eeR1 contains a possibly interesting InterruptibleLockedRandomAccessFile class.

[jamoore]

Splitting due to milestone decoupling.