- Infos im HLRS Wiki sind nicht rechtsverbindlich und ohne Gewähr -

Difference between revisions of "Laki"

From HLRS Platforms
Jump to navigationJump to search
Line 6: Line 6:
 
we would like to inform you how we plan to migrate the
 
we would like to inform you how we plan to migrate the
 
production from the old Cluster to Vulcan:
 
production from the old Cluster to Vulcan:
    - All active projects on Laki will automatically
+
* All active projects on Laki will automatically get access to Vulcan.
      get access to Vulcan.
+
* Login to Vulcan will be possible starting from October 1st  2018
    - Login to Vulcan will be possible starting
+
* The Lustre workspace filesystem and the users Home filesystems will be identical on both clusters (Laki and Vulcan).
      from October 1st  2018
+
* Compute nodes will be moved from Laki to Vulcan. The following time schedule is planned:
    - The Lustre workspace filesystem and the users Home
+
** 10/01-17: start moving special node types (graphics nodes, large memory nodes, ...)
      filesystems will be identical on both clusters
+
** 10/08 :start moving SandyBridge nodes to Vulcan
      (Laki and Vulcan).
+
** 10/10: start moving Haswell (hsw128gb10c) nodes to Vulcan
    - Compute nodes will be moved from Laki to Vulcan.
+
** 10/15: start moving Haswell (hsw128gb12c) nodes to Vulcan
      The following time schedule is planned:
+
* Final shutdown of Laki (all queued jobs on Laki will be deleted): October 31st
        * 10/01-17: start moving special node types
+
* The operating system on Vulcan will be based on CentOS 7.
                  (graphics nodes, large memory nodes, ...)
+
* The batch system will be [https://kb.hlrs.de/platforms/index.php/Batch_System_PBSPro_(vulcan) PBSPro].
        * 10/08 :start moving SandyBridge nodes to Vulcan
+
* accounting of the cluster nodes will be changed from core hours (Laki) to node hours (Vulcan).
        * 10/10: start moving Haswell (hsw128gb10c) nodes to Vulcan
 
        * 10/15: start moving Haswell (hsw128gb12c) nodes to Vulcan
 
    - Final shutdown of Laki
 
      (all queued jobs on Laki will be deleted): October 31st
 
    - The operating system on Vulcan will be based on CentOS 7.
 
    - The batch system will be [https://kb.hlrs.de/platforms/index.php/Batch_System_PBSPro_(vulcan) PBSPro].
 
    - accounting of the cluster nodes will be changed from core hours (Laki) to node hours (Vulcan).
 
 
}}
 
}}
  

Revision as of 12:02, 27 September 2018

Warning: FINAL SHUTDOWN of cluster laki: October 31st

Due to the outdated software version on the Laki cluster, and the upcoming installation of the new cluster named Vulcan, we would like to inform you how we plan to migrate the production from the old Cluster to Vulcan:

  • All active projects on Laki will automatically get access to Vulcan.
  • Login to Vulcan will be possible starting from October 1st 2018
  • The Lustre workspace filesystem and the users Home filesystems will be identical on both clusters (Laki and Vulcan).
  • Compute nodes will be moved from Laki to Vulcan. The following time schedule is planned:
    • 10/01-17: start moving special node types (graphics nodes, large memory nodes, ...)
    • 10/08 :start moving SandyBridge nodes to Vulcan
    • 10/10: start moving Haswell (hsw128gb10c) nodes to Vulcan
    • 10/15: start moving Haswell (hsw128gb12c) nodes to Vulcan
  • Final shutdown of Laki (all queued jobs on Laki will be deleted): October 31st
  • The operating system on Vulcan will be based on CentOS 7.
  • The batch system will be PBSPro.
  • accounting of the cluster nodes will be changed from core hours (Laki) to node hours (Vulcan).


The NEC Cluster is part of a heterogenious computing platform (Projekt name Baku). This Project has 2 independent PC-Clusters installed in 2 different buildings. The based parts are named Laki and Laki2.



Introduction


Troubleshooting



Cluster Documentation


Programming / Utilities



Help for Wiki Usage