Difference between revisions of "Euler maintenance (September 2017)"

From ScientificComputing
Jump to: navigation, search
m
(Updates)
 
(14 intermediate revisions by 4 users not shown)
Line 1: Line 1:
 
CSCS informed us that the data centre will be undergoing maintenance on '''Wednesday 27 and Thursday 28 September 2017''', which will require a '''complete shutdown''' of Euler.
 
CSCS informed us that the data centre will be undergoing maintenance on '''Wednesday 27 and Thursday 28 September 2017''', which will require a '''complete shutdown''' of Euler.
  
We will take this opportunity to do some additional maintenance on Euler itself the days '''before''' and '''after''' the CSCS maintenance. In particular, we will complete the '''upgrade of Euler to CentOS 7''' and '''prepare for the installation of Euler IV''' later this year.
+
We will take this opportunity to do some additional maintenance on Euler itself the days '''before''' and '''after''' the CSCS maintenance. In particular, we will complete the '''upgrade of Euler to CentOS 7''' and '''prepare for the installation of Euler IV''' later this year. The network at CSCS will also be updated on this occasion.
  
 
As a consequence we plan that:
 
As a consequence we plan that:
  
* Euler compute nodes, login nodes, and storage systems will be '''off-line from 12:00 (noon) Tuesday 26 until and including Friday 29 September'''; we will do out best to bring them back on-line ASAP.
+
* Euler compute nodes, login nodes, and storage systems will be '''off-line from 09:00, Tuesday 26 until the evening of Friday 29 September'''; we will do our best to bring them back on-line ASAP.
  
* Batch queues will be '''inactive throughout the weekend until Monday, 04 October 2017'''.
+
* Batch queues will remain '''inactive throughout the weekend until Monday 2 October 2017'''.
 
 
The exact times will be coordinated with CSCS and communicated here and by email in the weeks prior to the maintenance.
 
  
 
As usual, batch queues will be progressively inactivated in the days and hours prior to the maintenance, to ensure that no jobs get killed when the cluster is shut down.
 
As usual, batch queues will be progressively inactivated in the days and hours prior to the maintenance, to ensure that no jobs get killed when the cluster is shut down.
  
 
''Sorry for the inconvenience.''
 
''Sorry for the inconvenience.''
 +
 +
==Updates==
 +
 +
;'''2017-09-29 17:15'''
 +
 +
:'''The Euler cluster will remain unavailable until Tuesday 3 October 2017 to fix a critical security vulnerability disclosed during the maintenance ([https://access.redhat.com/security/vulnerabilities/3189592 CVE-2017-1000253])'''
 +
 +
;'''2017-10-02 16:20'''
 +
 +
:Our system administrators have '''fixed the security vulnerability''' in most of the nodes of the Euler cluster. We therefore decided that the '''login nodes of the Euler cluster will already be opened today'''. Users can login and access their files, but the '''queues stay inactivated''' due to further testing.
 +
 +
;'''2017-10-04 08:00'''
 +
 +
:We start to progressively open the batch queues starting with the 4h queue.
 +
 +
;'''2017-10-05 15:00'''
 +
 +
:All queues open again, Euler is back in production.

Latest revision as of 13:14, 5 October 2017

CSCS informed us that the data centre will be undergoing maintenance on Wednesday 27 and Thursday 28 September 2017, which will require a complete shutdown of Euler.

We will take this opportunity to do some additional maintenance on Euler itself the days before and after the CSCS maintenance. In particular, we will complete the upgrade of Euler to CentOS 7 and prepare for the installation of Euler IV later this year. The network at CSCS will also be updated on this occasion.

As a consequence we plan that:

  • Euler compute nodes, login nodes, and storage systems will be off-line from 09:00, Tuesday 26 until the evening of Friday 29 September; we will do our best to bring them back on-line ASAP.
  • Batch queues will remain inactive throughout the weekend until Monday 2 October 2017.

As usual, batch queues will be progressively inactivated in the days and hours prior to the maintenance, to ensure that no jobs get killed when the cluster is shut down.

Sorry for the inconvenience.

Updates

2017-09-29 17:15
The Euler cluster will remain unavailable until Tuesday 3 October 2017 to fix a critical security vulnerability disclosed during the maintenance (CVE-2017-1000253)
2017-10-02 16:20
Our system administrators have fixed the security vulnerability in most of the nodes of the Euler cluster. We therefore decided that the login nodes of the Euler cluster will already be opened today. Users can login and access their files, but the queues stay inactivated due to further testing.
2017-10-04 08:00
We start to progressively open the batch queues starting with the 4h queue.
2017-10-05 15:00
All queues open again, Euler is back in production.