Bug Alert for Dataverse 6.3 and prior releases

[James Myers]

Dear Dataverse enthusiasts,

 

There are two recently discovered significant bugs being fixed in v6.4. One, introduced in v6.3, can affect the layout/display of the main page. The other, while limited to functionality only available to superusers, can result in metadata loss and a broken dataset. Harvard Dataverse is running a patched version of v6.3 that we recommend for others as well. For earlier versions of Dataverse, which would be affected by the second bug, we recommend one of a few work-arounds. Details and workarounds are provided below.

 

– The Dataverse Core Team

• Publishing breaks designated Dataset thumbnail, messes up Collection page: When a file thumbnail is autoselected for a dataset and the dataset is published, a bug causes the default thumbnail to display with text from the title overlaying multiple display elements (see issue for a picture). Fixed in https://github.com/IQSS/dataverse/pull/10820 and included in the patched version linked above.

• “Update Current Version” can cause metadata loss: It was discovered that if the Update-Current-Version option is used when publishing a dataset version, and the PID provider is down or if any of the Terms fields have been changed in the new version, publishing will fail and result in the latest published version having no metadata. This results in a 500 error when attempting to view the dataset. Any affected dataset would need to be updated from a back-up copy. As this functionality is only accessible to superusers, and is most often used to edit metadata, we only know of one instance where the problem has been seen. 

 

This bug is fixed in https://github.com/IQSS/dataverse/pull/10797 which is included in the patched version linked above. The fix allows the update to succeed in both of these cases and redesigns the functionality such that any unknown issues should not make the dataset unusable (i.e. the error would be reported and the dataset would remain in its current state with the last-published version as it was and changes still in the draft version.)

 

Recommendations:

 

V6.3: Deploy the patched version of v6.3.

 

Prior versions, which are only affected by the Update-Current-Version bug:

 

Alert their superusers to this issue. Do not use Update Current Version to update any dataset terms and do not use it if there are known PID service or network issues. Make sure you have a recent dataset backup.

 

Those who wish to disable this functionality have two options:

1. Change the dataset.updateRelease entry in the Bundle.properties file (or local language version) to "Do Not Use" or similar (doesn't disable but alerts superusers to the issue)  and restart the Payara server, or

 

2. Edit the dataset.xhtml file to remove the lines

 

<c:if test="#{dataverseSession.user.isSuperuser()}">

  <f:selectItem rendered="#" itemLabel="#{bundle['dataset.updateRelease']}" itemValue="3" />

</c:if>

 

After editing the file, delete the contents of the ./generated and ./osgi-cache directories in the Dataverse Payara domain, and restart the Payara server.

 

[Clemens, Vera]

Hi,

thank you for the patched version!

We are using the Docker images (https://hub.docker.com/r/gdcc/dataverse) for our deployment. Is there or will there be a (tagged) image released for the patched version?

Best regards,

Vera

On Tue, 2024-09-10 at 22:40 +0000, James Myers wrote:

Dear Dataverse enthusiasts,

 

There are two recently discovered significant bugs being fixed in v6.4. One, introduced in v6.3, can affect the layout/display of the main page. The other, while limited to functionality only available to superusers, can result in metadata loss and a broken dataset. Harvard Dataverse is running a patched version of v6.3 that we recommend for others as well. For earlier versions of Dataverse, which would be affected by the second bug, we recommend one of a few work-arounds. Details and workarounds are provided below.

 

– The Dataverse Core Team

• Publishing breaks designated Dataset thumbnail, messes up Collection page: When a file thumbnail is autoselected for a dataset and the dataset is published, a bug causes the default thumbnail to display with text from the title overlaying multiple display elements (see issue for a picture). Fixed inhttps://github.com/IQSS/dataverse/pull/10820 and included in the patched version linked above.

[Eryk Kulikowski]

Hi,

Patching v6.3 docker image is more tricky since the base image (JDK) has changed and the shell scripts from v6.3 do not work anymore. FWIW, we build the images from source and there is an updated build script that does build the patched image on our GitHub repo: https://github.com/libis/rdm-build

The script to build the patched version: https://github.com/libis/rdm-build/blob/main/images/dataverse/build_dv.sh (you will need to define the env variables, or replace them with values, and make sure the correct versions of maven and JDK are installed, see also https://github.com/libis/rdm-build/blob/main/.tool-versions -> you can use "asdf install", if you are using asdf).

Hope that helps.

Greetings,

Eryk

Op woensdag 11 september 2024 om 11:19:20 UTC+2 schreef cle...@zbmed.de:

[Clemens, Vera]

I see, thanks Eryk!

[Philip Durbin]

Yes, thanks Eryk. I went ahead and added this topic of patched Docker images to our agenda for tomorrow: https://ct.gdcc.io

My first thought is that when we make a proper release (6.3.1 in this case), the fixes make it into the "alpha" image, which tracks the "master" branch.

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-commu...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/ec443d42bd26ef4abda118437b7262ca7f53c04b.camel%40zbmed.de.

--

Philip Durbin
Software Developer for http://dataverse.org
http://www.iq.harvard.edu/people/philip-durbin

[Philipp Conzett]

Hi Phil and Eryk,

Thanks for the heads-up!

Could you provide an estimate of when a fixed release of the dockerized version (6.3.1) will be available?

Best,

Philipp

[Philip Durbin]

Philipp, thanks for joining the Containerization Working Group meeting this morning to discuss this. I'll summarize the discussion, but if anyone would like to watch it (the first topic), please find the recording at https://harvard.zoom.us/rec/share/k7-UbJ-mx_fWANw-8wzt5v6xaRWH7_Fh48cqGDv2XVOqxszFEvNnVhfECabtlk2M.6e5v_FFHjBwIPjAF

No, we don't plan to release Dataverse 6.3.1. As Ceilyn indicated* we intend to release Dataverse 6.4 soon, possibly as even in a couple weeks.

We should probably be more conscious of Docker users. In the future, when the team starts rumbling about the need to put out a patch, I'll try to remind everyone that our Docker images are only updated when we do a regular release.

As Eryk points out above, it's possible for anyone to create a Docker image that includes an updated war file, but this is an exercise for the reader. Perhaps Eryk's scripts could be used as a starting point.

I hope this helps. Again, there's more discussion in the recording above. Onward to Dataverse 6.4!

Thanks,

Phil

* https://groups.google.com/g/dataverse-community/c/EV_JJGMaEKg/m/rYOEw3GnCAAJ

To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/9f0d6f07-d467-44f5-bc5d-083a642f50e4n%40googlegroups.com.

[Philipp Conzett]

Thanks for clarifying, Phil!

Best,

Philipp

[Weihong Xu (link89)]

For the one that use this docker solution:

link89/dataverse-docker: Towards production ready dataverse docker deployment. (github.com)

the fix can be applied with the following way:

1. copy patched war into the dvinstall folder to replace the 6.3 war.

2. restart docker (compose) 

[meghan.good...@gmail.com]

Hello!

Do you know if there is any documnetation for how to restore a dataset from a back-up copy. In an attempt to fix some datasets with an "incomplete metadata tag", we have experienced this bug as well.

Thanks for any help you may be able to offer!

Best,

Meghan

[Alfredo Cosco]

Hi all, I'm experiencing this issue on a 6.3 installation, 

now I have a dataset updated with all metadata missing, and I need:
1) a quick fix: Is it possible to go back to the previous version?

2) a medium fix: i will disable the functionality as explained at the beginning of this thread. But then, is it still possible to add a new version of a dataset? 

3) a long term fix: How to update to the patched version? Do I have just to undeploy the old and deploy the new one?

Thanks,

Alfredo