Fwd: URGENT: SCS Compute Clusters - reboot required

[Edward Walter]

FYI... we will be rebooting the coma cluster on Thursday February 18th
between 8Am and Noon.

Thank you.

-Ed Walter
SCS Computing Facilities

-------- Forwarded Message --------
Subject: URGENT: SCS Compute Clusters - reboot required
Date: Wed, 17 Feb 2016 15:50:31 -0500
From: Edward Walter <ewa...@cs.cmu.edu>
To: Help Desk <He...@cs.cmu.edu>

February 17, 2016

A critical, remotely-exploitable vulnerability that affects most Linux
hosts was announced on February 16, 2016. This vulnerability affects
most of the SCS supported ROCKS based clusters as well. The list of
affected clusters is included below.

In response to this vulnerability, SCS Computing has installed updated
software on all of the systems in the high performance compute
clusters. To enable the security fix, every system in the each
cluster must be rebooted.

We plan to begin rebooting cluster servers (including head nodes,
login nodes, NAS servers, and web servers) Thursday February 18th
between 8 AM and Noon. We will reboot compute nodes opportunistically as
existing jobs complete so that long running jobs are not disrupted.
Unpatched compute nodes will be removed from the queues until they have
been updated.

We're notifying cluster owners about this update. We are also posting
an alert on the clusters themselves. Please share this notice with the
appropriate people in your labs or research groups.

If, for any reason, we can NOT reboot your cluster; please contact
help@cs and/or Edward Walter <ewa...@cs.cmu.edu> as soon as possible.


Affected Clusters
----------
actr
boston-cluster
cab
cogito
coma
cortex
homeuse
lanec1
latedays
leonid
mu (non-rocks cluster)
psych-o
rocks.is
steamroller
tuborg
workhorse
yoda


Additional information
----------
Technical details about this issue:
https://access.redhat.com/security/cve/cve-2015-7547

Please contact he...@cs.cmu.edu or call the SCS Help Desk (x8-4231) if
you have questions or concerns about this issue.

Thank you for your attention,
SCS Help Desk

[Shirley Ho]

Hello All,

Please note that the Lustre filesystems are going away soon !!

Please make appropriate planning.

Shirley


>>>> As previously discussed, we plan to switch the older lustre
>>>> filesystems (/lustre /physics) on warp and coma to read only mode on
>>>> Tuesday March 1st. We've seen >5 disk failures on these filesystems
>>>> since December 2015. We're concerned that this pattern will continue.
>>>>
>>>> Unless we hear otherwise, we'll plan on going forward with these
>>>> changes. The Graphics users are planning for this as well.

> Here's the method I would use though in moving things to different storage:
>
> * historic/archival data -> move to NAS storage
> * active data that doesn't require fast I/O -> move to NAS storage
> * active data requiring fast/parallel I/O -> move to lustre (/physics2)