Opened 9 years ago

Closed 9 years ago

#724 closed help (fixed)

changes to pathscale compiler

Reported by: luke Owned by: um_support
Component: HECToR Keywords: pathscale,GCOM
Cc: Platform:
UM Version: 7.3

Description (last modified by ros)

People in Cambridge have been having problems with the pathscale compiler. Currently most peoples settings are to load the pathscale/3.2.99 compiler, but this seems to have been withdrawn as of the maintenance yesterday. Looking at the new compiler should be pathscale/4.0.9 (as is the same with the module avail command).

It may be that our .profile settings need updating. I currently have TARGET_MC=pathscale_quad, and this points to /work/n02/n02/hum/vn7.3/pathscale_quad/scripts/.umsetvars_7.3. However, when logging into HECToR I get the error

PathScale PrgEnv loaded
pathscale(3):ERROR:105: Unable to locate a modulefile for 'pathscale/3.2.99'
ModuleCmd_Switch.c(172):ERROR:152: Module 'xt-mpt/5.3.1' is currently not loaded

so although it is loading 4.0.9, it is having problems loading some things.

When compiling the UM, it give the following error with gcom:

 Module file "/work/n02/n02/hum/gcom/pathscale_quad/gcom3.4/hector_pa_mpp/inc/MPL.mod" is incompati
ble with this compiling system.  Recompile the module with this compiling system.

Does this need recompiling with 4.0.9, or do we need different settings in the .profile?

This has affected all UM users in the group.


Change History (6)

comment:1 Changed 9 years ago by grenville

Hi Luke

I have been in contact with Cray at HECToR in relation to the problems compiling today. Regrettably, I think we will just have to stop developing until phase3 is in place. When they upgraded the OS yesterday, they removed the pathscale compiler that we normally use (this was not made clear in any communication I had from them). I have enquired about the possibility of reinstating it, but they won't (can't) do that.

We could try to use the new Cray compiler (the one we have been testing on the tds, but that'd mean rebuilding gcom), and that would only help for a couple of days before the disruption next week.

Binaries built yesterday should still run fine with the new OS.

Can you inform your group of this. I'll be happy to if not.



comment:2 Changed 9 years ago by luke

Hi Grenville,

I'll let the group know.

Do you have a timetable of the upcoming disruption so that I can let people know how long it is likely to last?



comment:3 Changed 9 years ago by grenville


All Hector users should have got this email - I hope they did anyway:

Main Interlagos Upgrade Timeline

The upgrade to Interlagos processors from Magny-Cours will take place in two parts,

and will require 3 maintenance sessions. During the first part half the current

number of Magny-Cours processors will be available to users (896 nodes; 21,504 cores).

During the second part, twenty cabinets of newly installed Interlagos processors
will be available (1856 nodes, 59,392 cores). Finally, HECToR will be returned as
a 30 cabinet Interlagos system (2816 nodes, 90,112 cores).

This timeline will be updated as and when there are any changes to the plan. We will

notify users as soon as possible if the dates below are expected to change.

Outage 1: Mon 07 Nov 0900-1700

HECToR will be restored at 50% capacity e.g. a 10-cabinet Magny-Cours system.

Outage 2: Weds 9 Nov 0900 - Friday 11 Nov 0900

HECToR will be restored as a 20-cabinet Interlagos system.

At this stage user accounting will be disabled. Jobs will be recorded but you will

not be charged. Accounting will remain disabled until such time that the acceptance
testing is completed (estimated early December).

Outage 3: Weds 16 Nov 0900 - Thurs 17 Nov 0900

HECToR will be restored as a 30 cabinet Interlagos system. User accounting remains

disabled at this stage. Acceptance and availability testing is scheduled for late
November/early December.

Details for preparing for Interlagos will be published at a later date.
Any users requiring assistance should contact the HECToR Helpdesk.


comment:4 Changed 9 years ago by luke

Hi Grenville,

Thanks for that - I did have it, but I was wondering if you had any more information over the standard email.


comment:5 Changed 9 years ago by grenville


Nothing from Hector. I am writing a web page to indicate what users will need to do next Friday and will send a message to all n02 users pointing them to the page. With a bit of luck users shouldn't need to do much to get running again.



comment:6 Changed 9 years ago by ros

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