Nice to see a new project, Mapping Arthritis Markers.
Only one problem. downloaded some cancer markers (successfully) then it downloaded 101 arthritis markers all failed.
Coms backed off 24 hrs.

Edit. Downloads were at 11:15(BST) this morning.

In reply to Sirius B's message of 13 May 2026:

Nice to see a new project, Mapping Arthritis Markers.
Only one problem. downloaded some cancer markers (successfully) then it downloaded 101 arthritis markers all failed.
Coms backed off 24 hrs.

Edit. Downloads were at 11:15(BST) this morning.

I think that you need to install one of the newer Visual C++ Redistributable 2015-2022, or even Visual C++ Redistributable 2022, to make MAM work. I just installed Visual C++ Redistributable 2015-2022 on one of my Windows 8.1 computers. Maybe it works for MAM on Windows 8.1. I think that Visual C++ Redistributable 2022 package is too new for Windows 8.1.

See the following part of the Operational Status page, from January 7, 2026:

Windows build of MDMG/MAM1 v7.08 successfully completed, deployment to beta30 in progress. We may have to include a large number of DLLs including pre-built LibTorch CPU libraries at first. These range in system requirements for Windows 10/11 64-bit with Visual C++ Redistributable 2015-2022, but for this initial build that successfully ran the test suite for the LibTorch backend in powershell on Windows 10, we expect that Visual C++ Redistributable 2022 is required. We've documented all third-party licenses (all permissive: BSD, MIT, Apache, NCSA, Zlib), and are hopeful a full static build will follow. System requirements are expected to be 4GB RAM or more, about ~300MB disk for application + DLLs, and we will be running the Windows build exclusively in beta for now starting sometime this week, and ramping up with the Linux build as we analyze first results.

WCG forum is down:

mvnForum Fatal Error Message : Cannot init system. Reason : Assertion in ForumUserServlet.

popping in to say hi here, since WCG forums are down (as Grumpy Swede already said)

Hello !

Update from https://www.cs.toronto.edu/~juris/jlab/wcg.html
(WCG Forum still down)
May 18, 2026
Recent MAM1/beta30 smoke testing batches unbounded memory use on BOINC clients - the cause was a bug in the dataset loader, combined with a placeholder dataset file that was pushed from staging to prod with the batch generation logic. In place of the correctly structured and non-empty MAM1 dataset, this smaller file being read by the dataloader without guards against the invalid formatting caused the OOM crashes reported by users. We deprecated the application and cancelled all workunits when we saw this happening, and we have fixed the issue in the dataset loader in a new build of the MAM1 application. We then released a handful (10) beta30 project workunits tonight May 15th, 2026 to confirm the fix, and we will resume smoke testing once we have Windows, WSL, and Docker support tested through the beta30 project. Expect the beta testing thread about MAM1 next week, preceeding any further smoke testing in the MAM1_9999900+ batch range to exercise the production lifecycle. We apologize for the inconvenience, we did test locally and in our staging environment but did not catch this as with the correct dataset, it worked.
Results API not showing IN_PROGRESS, not consistent with authoritative BOINC database - we will be switching out the connection to the legacy database that currently serves the Results API for a connection to the new postgres cluster coordinator, so that the website will then fetch authoritative data from postgres including IN_PROGRESS results. When this is ready, it should resolve many of the mysterious missing or inconsistent states for results reported by volunteers in the forum.
Found the bug in the file_upload_handler causing the missing validations issue generally for MCM1, affecting stats and results - working on the fix after sweeping the database and file system to find leftover inflated credit values and issues resulting from the outages at the data center, partition by partition.

---

Paul.

Forums still down with same message

mvnForum Fatal Error Message : Cannot init system. Reason : Assertion in ForumUserServlet.

Forum still down, BOINC & other web pages appear OK.

---

Paul.

In reply to PMH_UK's message of 19 May 2026:

Forum still down, BOINC & other web pages appear OK.

Well, with the forum down, the team doesn't have to bother about all the complaints that people post. They already know that the system is in a mess. :-)

Seemed weird to me that they posted an update, but didn't mention the forum being down.

I sent an email to support.

Grumpy Swede the forums aren't just for complaining. It is also (lately) to wildly suggest they code up something some other site has done even though they don't have a stable system.

I will give the WCG techs kudos on getting the flow of WUs stable. I'm still getting mostly resends. I do hope they extend the deadline. Maybe they have already extended the deadline. Something for us to discuss when we get our forums back.

I am getting about half resends. The perhaps worrying thing is, those resends seem to have a high percentage of _3 or later sends of tasks that implies either problems with tasks returned to the upload server, something causing a higher than normal error rate, or a lot of tasks timing out because of problems somewhere in the system.

May not be erroring. I am getting so many that the client doesn't have enough time to finish them all.

Forum now up, no new posts yet.

---

Paul.

In reply to just1vet's message of 19 May 2026:

May not be erroring. I am getting so many that the client doesn't have enough time to finish them all.

Odd, I thought BOINC was supposed to stop that happening? On my box, it estimates tasks will take just over 2.5hours whereas they actually take about 50 minutes. Mind you I have my cache set very low.

WCG forum is back !!

In reply to Dave's message of 19 May 2026:

Odd, I thought BOINC was supposed to stop that happening? On my box, it estimates tasks will take just over 2.5hours whereas they actually take about 50 minutes

On one computer, probably the fastest, the estimate is 1:03 and actual runtime seems to be about 1:07. Other computers have different estimated times up to 3:38, actual about 4:50, for the slowest. I keep my cache set to 2d +.01 on all of them, except for the one which is away from home 3 days a week with me (no internet access so it's set for 3d).

I did wonder if it was to do with benchmarks and I ran the benchmarks again. Over 40 tasks completed since and the estimated time to completion on new tasks has not changed. BOINC8.0.3, Ubuntu 26.04 Last CPDN task I ran before they ran out was pretty close to the 53 hours for a task to complete. There is still an open issue over at git-hub for this. When I saw the nearly correct figure for CPDN I thought maybe it had been sorted b but clearly not.

New Interesting? Error message, when trying to reach the forum:

503: Load Balancer is unable to route your request to an available server, most likely there is an issue in our infra that we are working to resolve.
BOINC services and the webserver are different backends, so one may be still be handling traffic while the other is down, and in the case of the BOINC backend it is also possible the servers are temporarily too busy to handle your request at this time, and trying again shortly will work as normal.

WCG Operational Status on the Jurisica Lab Website
For further details about recently completed work, current priorities, and known issues, see the Operational Status tab (top right of the nav bar on the page linked above).

Edit: That was a very short error. The forum is reachable again.