Closed jwodder closed 4 weeks ago
So it seems this is mainly a retroactive issue? That is, we should update these respective versions to reflect the last modified time of those assets, but it's not a systematic issue going forward? Seems like it's just a side effect of the embargo re-design data migration.
@jjnesbitt Yes, that sounds correct.
@jwodder I believe I've addressed the issue. Can you verify this?
@yarikoptic ping
Thank you @jjnesbitt ! FTR: I have triggered rerun of flock -E 0 -e -n /home/dandi/.run/backup2datalad-cron-nonzarr.lock bash -c '/mnt/backup/dandi/dandisets/tools/backups2datalad-update-cron --mode verify
so would have an answer (hopefully) when it completes.
@jjnesbitt do you recall any details on what you have done since no detail were provided? I posted the summary of current situation to https://github.com/dandi/backups2datalad/issues/49#issuecomment-2402911554 and it seems that 000874 also has odd modified date which is before when updated assets were uploaded -- not sure if it is exactly this issue or something else, please have a look if you just went for a single initially reported dandiset.
@jjnesbitt do you recall any details on what you have done since no detail were provided?
I did what I said I would in this comment, which is, updated all of the relevant draft versions to reflect the last modified time of their modified assets. However, I only did so for assets with a last modified time of April 29, as that was the day of the data migration that John filed this issue about, and I don't want to accidentally modify data incorrectly.
Since this issue is specifically about assets related to the embargo data migration, I'd like to close this issue as resolved. If there are more instances of this occurring, we can investigate them separately.
I posted the summary of current situation to https://github.com/dandi/backups2datalad/issues/49#issuecomment-2402911554 and it seems that 000874 also has odd modified date which is before when updated assets were uploaded
I'm not sure the extent of affected dandisets, as it's fairly hard to make sense of that comment with all of the data. Is it just that dandiset? Do you know how many assets are affected?
Quick one: that type of error indeed was only in one dandiset. It was under a dozen assets. I will later return to get a full list of that would help.
In that case I think we can close this issue. If/when you have that data for the other dandiset, it could be filed as a separate issue.
It appears that, on or around April 29, embargoed asset blobs were moved from the
dandiarchive-embargo
bucket to thedandiarchive
bucket. This was accompanied by updates to the assets'contentUrl
metadata, and the assets'modified
properties were updated at that time, but themodified
properties of the respective Dandisets'draft
versions were not updated, andbackups2datalad
does not like this.For a specific example, the embargoed Dandiset 000770 contains an asset
sub-Elgar/sub-Elgar_ses-2022-06-04_ecephys.nwb
with amodified
property of2024-04-29T19:35:05.558024Z
, yet themodified
property for 000770/draft is still2024-04-09T19:15:02.126137Z
.