Open Microscopy Environment

After importing images via OMERO.insight-5.4.6-ice36-b87-linux, and browsing them in OMERO web (5.4.6-ice36-b87) I observe something weird:
Thumbmail images are different (ie not the same images/fields) from those displayed in the viewer (or when double-clicking to enlarge the image in another tab), as seen in attached screenshot.
Any idea what is wrong?

NB: This happens only with the last imported images, not with older imports.

Dear Philippe,

That's alarming; I don't think we know of any other such reports. Your screenshot is from simply browsing a dataset? Are there any UI actions (searching, filtering, whatever) that make things start to appear wrongly, or does the newly imported data have thumbnail-to-full-view linking wrong at all times?

So, the preview window on the right is showing you the same as what you get in the full viewer, just not matching the thumbnail?

Have you tried in OMERO.insight? I wonder what happens in that for you on this data.

Cheers,
Mark

Yes Mark, that's just browsing my last Insight import on the OMERO web interface. No searching, filtering, or anything else.
And yes, every time I tried to open an image via its thumbnail, I got a wrong image. Again, this is not true with earlier data sets.
Same exact behavior when working through Insight on these last imported datasets.
One "reassuring" thing is that the images obtained from a specific thumbnail via either OMERO web or Insight are both the same image, even though not the one from the thumbnail.
I hope I am clear enough, here...

Dear Philippe,

Your answers are indeed clear, thank you; what an interesting problem you're witnessing. Let's look for further clues.

Can you tell, by image name or whatever, whether it's the thumbnail or the full image that's wrong? Do you notice any relationship between or pattern with the correct images and the wrong? (E.g.: Are the wrong ones all recycled from old imports?)

You've checked that nobody knows of any other sysadmin changes that happened -- to the server or to its underlying infrastructure / dependencies -- at around the time the problem began to present?

Cheers,
Mark

The thumbnails are wrong, but the full images (from preview and/or double clicking the thumbnails) are the expected ones.

The thumbnails indeed do come from other datasets (but I could not identify the source: too many of them...).

And yes, we recently had a change on our server: Upgrade from 5.2 -> 5.3 -> 5.4.6, and data recovery from our backup. No problem before these upgrades.

Dear Philippe,

Aha, so you are seeing the wrong thumbnails. Does this thumbnail issue coincide with any -- perhaps all -- of these upgrade / recovery events? There have certainly been some thumbnail service changes since 5.2 but I am also trying to figure if some oddness in the data recovery could cause this. You've not noticed any other strange new regressions, right, it's only this thumbnail one?

Cheers,
Mark

Hi Mark,
We did the 5.2 ->5.3->5.4 without using 5.3, just as an intermediate step in the 5.2 to 5.4 upgrade.
And we have not noticed anything wrong since the upgrade yet, except for this thumbnail issue.

Hi PhilippeP

Can you give us more information (as much as possible) on the steps that were taken during the the data recovery process. For example, did you recover the entire disk, or just some directories, and if so which ones. Were any changes made to the system between the last backup and the restoration steps?

Dear Philippe,

Also, from the psql shell for your OMERO database could you try these two queries and let us know the result from each?

Code: Select all

SELECT id FROM thumbnail ORDER BY id DESC LIMIT 1;
SELECT ome_nextval('seq_thumbnail');


Ideally the second should give a number a little higher than the first.

Cheers,
Mark

Oops...
I'm afraid this is beyond my biologist computing abilities. I could not even start a "psql shell for your OMERO database" event though I do have the server here in my office with all accesses...

I'll try to reach the computer person who did that for us, if he is not on vacation yet.
I'll keep you posted when I know more.

Thank you anyway for looking into the problem so quickly!
Philippe