Opened 4 years ago

Closed 4 years ago

#1689 closed help (fixed)

xconv still not working!

Reported by: charlie Owned by: um_support
Component: ARCHER Keywords:
Cc: Platform:
UM Version:

Description (last modified by ros)

Hi,

Despite the connection between Archer and Reading being fixed, I'm still having problems using xconv. It loads fine, but when I try to open a file I either get the following error:

eslogin008:cjrw09$ xconv xlmhga.dak1110
Segmentation fault

or, alternatively, lines and lines of error messages.

This only seems to happen when I open a start dump i.e. *.da*. I have tried it with multiple files, and its the same for all dumps. Moreover, this problem is not specific to my account.

Please would you investigate as a matter of urgency?

Thanks,

Charlie

Attachments (1)

xconv_errors_1650hrs081015.txt (12.1 KB) - added by charlie 4 years ago.

Download all attachments as: .zip

Change History (10)

comment:1 Changed 4 years ago by charlie

PS. This problem may not actually be related to Archer, as I have just tried sftp-ing a couple of the problem files to our system, and I'm getting the same problem with xconv

comment:2 Changed 4 years ago by charlie

PPS. Although better than it was, the connection with Archer is still very slow. I'm only getting transfer speeds of around 5 MB/sec, which is much slower than usual

comment:3 Changed 4 years ago by andy

Hi Charlie,
Please try to be more specific when reporting problems. If you give us concrete examples of things and details of what machines and files you're transferring and at what time then we can look at the issue. Saying "I have a problem" without specifics gives us very little to go on. In this case it could be and Archer problem, a network problem or a local server / disk problem.

In this case for example you could have written:

I'm still experiencing a slow network connection from login.archer.ac.uk and jasmin1.rdg.ac.uk. The command I'm doing is:

scp ajh@…:/nerc/n02/n02/ajh/idl.pdf /panfs/jasmin/users/andy

At 12:40pm today I got a transfer rate of 5MB/sec and at 14:23pm I got a transfer rate of 3MB/sec.


This sort of email is much easier for us to look at your problem and has the information we need to reproduce it.

I ran this command at 14:49 today and got a transfer rate of 32MB/sec.

Cheers
Andy

comment:4 Changed 4 years ago by charlie

Very well. To repeat my earlier message, with specifics:

I am connecting to login.archer.ac.uk with username cjrw09. Using xconv v1.93, I am trying to open any of the .da files (which are in standard UM format, produced by UM v6.6.3) in /work/n02/n02/cjrw09/result. There are 4 .da files in this directory:

eslogin005:cjrw09$ ls -lh *da*
-rw-r--r-- 1 cjrw09 n02 1.8G Sep  8 06:54 xlmhea.dak1110
-rw-r--r-- 1 cjrw09 n02 1.8G Sep  8 07:05 xlmhfa.dak1110
-rw-r--r-- 1 cjrw09 n02 1.8G Sep 11 23:27 xlmhga.dak1110
-rw-r--r-- 1 cjrw09 n02 1.8G Sep 17 02:56 xlmhha.dak1110

xconv loads fine, but when I try to open a file using the interface I get a load of errors (see attached). Alternatively, if I try to open a file from the command line, I get the following error:

eslogin005:cjrw09$ xconv xlmhea.dak1110
Segmentation fault

All of the above occurred at 1650 hrs BST on Thursday 8 October 2015.

This only seems to happen with .da files, and xconv appears to load other files (e.g. .pa) with no problems. This is also true for files on the RDF e.g. any .da files in /nerc/n02/n02/cjrw09/hydro.d/exp_b1.d/xlmha.d do not load

Moreover, this problem is not specific to my account.

This problem may not be specific to login.archer.ac.uk as this morning I also transferred 2 of the above .da files to my home directory /home/charlie on jasmin1.rdg.ac.uk and exactly the same problem occurs.

The transfer (using sftp) speed was significantly slower than usual, at 4.9-5.1 MB/sec. I can't remember exactly what time this was, but it was around 1000 hrs BST on Thursday 8 October 2015. This issue, however, is less of a problem than the xconv issue.

Thanks,

Charlie

Last edited 4 years ago by ros (previous) (diff)

Changed 4 years ago by charlie

comment:5 Changed 4 years ago by jeff

Hi Charlie

This is a known xconv problem with certain dump files. There are some ways of specifying STASH diagnostics which can confuse xconv. I've just released a new version of xconv and this crashes with your files, the previous version doesn't crash, although it has exactly the same problem with certain files. As a temporary workaround you can run xconv1.92, this seems to work ok on file xlmhga.dak1110 but may still crash on other files. I will look into trying to stop xconv crashing.

Jeff.

comment:6 Changed 4 years ago by andy

Hi Charlie,

Thanks for the detail - it really helps to see what's going on.

Copying data to your home directory is much slower than copying to data disks. I would recommend using jasmin1 - jasmin6 as they are on a 10Gb/sec network and are our fastest link in and out of the department.

This morning I copied that file from Archer to jasmin2 using scp and it came across at 79MB/sec.

Thanks
Andy

comment:7 Changed 4 years ago by ros

  • Description modified (diff)
  • UM Version <select version> deleted

comment:8 Changed 4 years ago by jeff

Hi Charlie

I've modified the latest version of xconv so it doesn't crash on your dump files. You should now be able to use the default version (1.93).

Jeff.

comment:9 Changed 4 years ago by jeff

  • Resolution set to fixed
  • Status changed from new to closed
Note: See TracTickets for help on using tickets.