You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: mkdocs-project-dir/docs/Status_page.md
+72-11Lines changed: 72 additions & 11 deletions
Original file line number
Diff line number
Diff line change
@@ -779,15 +779,76 @@ This page outlines that status of each of the machines managed by the Research C
779
779
780
780
**Removal of old filesystem**
781
781
782
-
`/old_lustre` will be available for three months, until 9am on Monday 7 July. It will then be unmounted and you will not be able to access it any longer.
782
+
`/old_lustre` will be available for three months, until 9am on Monday 7 July. It will then be
783
+
unmounted and you will not be able to access it any longer.
783
784
784
785
**Myriad at risk for first week**
785
786
786
-
Myriad should be considered exposed to potential issues for the first week of running a full workload with the new filesystem, and so there might be interruptions to service if anything goes wrong or needs tuning differently.
787
+
Myriad should be considered exposed to potential issues for the first week of running a full
788
+
workload with the new filesystem, and so there might be interruptions to service if anything goes
789
+
wrong or needs tuning differently.
787
790
788
791
The new filesystem is GPFS (IBM Storage Scale) and not Lustre, for those who are interested.
789
792
790
-
Additional FAQs will be added here based on questions we receive.
793
+
Additional FAQs will be added here based on questions we receive.
794
+
795
+
- 2025-04-14 - **Myriad filesystem update and issues with symlinks**
796
+
797
+
This is a quick rundown of what else happened on Myriad last week and then some tips for problems
798
+
people have been having.
799
+
800
+
After the new filesystem went live, we had a few issues on Wednesday and Thursday where some jobs
801
+
were causing nodes to crash which was in turn causing the gpfs client to hang - which you will have
802
+
seen as timeouts or very slow access on the login nodes. The hangs also meant that a few people had
803
+
their new home directories only half-created, so didn't have a home directory that belonged to them
804
+
when they logged in. We changed some configuration on the compute nodes to fix the issue (the jobs
805
+
causing the problem were running out of virtual memory). People who had the home directory issue
806
+
should have been sorted out on Thursday and Friday - let us know if anyone else still gets an error
807
+
about their home directory not existing.
808
+
809
+
We were running more smoothly by Friday. Issues like these are why we said the rest of that week was
810
+
at risk, as there was likely to be something that needed adjusting once real jobs started.
811
+
812
+
**Symbolic links and Scratch**
813
+
814
+
You start out with an empty normal directory called Scratch in your home. What I had not considered
815
+
is if you rsync the whole of your oldhome back in, then it will rsync the old Scratch symlink
816
+
(shortcut) from oldhome and replace the empty Scratch directory with it. This only happens because
817
+
that directory is empty.
818
+
819
+
We have had tickets from some of you about finding that files are read-only that you think you have
820
+
copied - it is because they are still really on the old filesystem.
821
+
822
+
If you do an ls -al in your home you will be able to see if you have ended up with something similar
0 commit comments