Internal Server Error on dataset page after deleting files, Vanishing files from already published versions. #8867
Comments
|
I have now commented out the corresponding img tag in dataset-citation.xhtml. This allows me to display the dataset. In doing so, I found that the files were indeed deleted, but directly in the published version V8. No draft version was created automatically :-( |
|
Correction: There is a DRAFT version but the files have been deleted in the published version V 8.0
|
lmaylein
changed the title
|
For the six files in the screenshot above, it is noticeable that the owner_id in dvobject is empty. In the other files of this dataset the owner_id is assigned. |
|
We have spent some time trying to diagnose this issue today; this appears to be a very serious problem, and we are working to produce a fix asap.
Could you please clarify what's going on. So, the dvobjects are still there - but the owner_id got nulled. This is consistent with what we are seeing in the code that was added in 5.11. That can of course be fixed - by setting owner_id=[dataset id]. Unfortunately, in addition to these problems on the database side, there is also a chance that the physical files for these 6 datafiles have been erased, from the filesystem/S3, whichever you are using. Could you please check your storage and confirm this? This of course would be the worst part, so this is why we are trying to fix this issue asap. |
|
Hello, |
|
This is, how it looks like in dvobject: And yes, it seems that the files were erased in files/studies/10.11588/data/DVU14P. |
|
Can I get the files back from backup and set the owner_id in dvobjects manually to possibly fix the problem or should I keep the current state for further debugging? |
|
Yes, I believe resetting the owner_id in these dvobjects, and restoring the physical files from backups will fix this. |
Yes, it works. I was also able to undo the changes in dataset-citation.xhtml. Thanks. |
|
Glad to hear it worked. This is a seriously destructive bug (yours was not the worst case). Sorry you had to be the first known victim of it, but thank you for bringing it to our attention. |
|
I have applied the patch version now, although we have only a few data curators who maintain the files and everyone is warned. |
|
The 5.11.1 release notes say: "if you delete a file from a published version of a dataset that has restricted files, the file will be deleted from the file system". |
|
Good catch! That statement is probably a hold-over from when we thought a dynamically generated draft version was involved and that only occurs when restricted files exist. (The dynamic draft is involved in the other issue and hence that one is limited to datasets with restricted files.) |
|
I continued to edit the dataset and deleted more files. The error did not reoccur with 5.11.1. Thanks! |
@lmaylein thanks, we updated the release notes and blog post. Much appreciated! |
Calling https://heidata.uni-heidelberg.de/dataset.xhtml?persistentId=doi:10.11588/data/DVU14P results in an Internal Server Error.
I have deleted files from this dataset (as I did many times before) to replace them in the resulting draft version and republish the dataset with the replaced files. Already when deleting the files I got an error message. Now the dataset is no longer accessible at all (in any version).
How can I make the dataset accessible again?
server.log:
The text was updated successfully, but these errors were encountered: