how to update from centos 7.2 to 7.6 in a cluster

General support questions
Post Reply
luiscssz
Posts: 7
Joined: 2020/03/26 18:42:04

how to update from centos 7.2 to 7.6 in a cluster

Post by luiscssz » 2020/03/26 20:36:16

Hi All,

my scope: update the cluster of my chair (8 nodes+master= 172 cpus) from centos 7.2 to centos 7.6 + all dependencies and not to the last available version (8).

reason: I have to install a specific version of ANSYS 19R2(multiphysics simulation software), which only works under 7.6.

I wanted to use the following command:
yum update (see doc)

but I am not sure if the release would be updated autmatically from 7.2 to 8. Do you have any experience with?

Already tried without success:

downloaded the Everything.iso image from here . After validation of the image file using the following procedure (see following 3 links: link1 , link2 and link3 ), the next issue appears: I cannot mount the .iso file (using the following procedure ), using:
$mount -o -loop CentOS-7-x86_64-Everything-1810.iso /media/7_6_centos/

mount: /dev/loop0 is write-protected, mounting read-only
mount: wrong fs type, bad option, bad superblock on /dev/loop0,
missing codepage or helper program, or other error
In some cases useful info is found in syslog - try
dmesg | tail or so.

Any help would be welcome because I have the cluster stopped

Thanks in advance.

Luis

User avatar
jlehtone
Posts: 4531
Joined: 2007/12/11 08:17:33
Location: Finland

Re: how to update from centos 7.2 to 7.6 in a cluster

Post by jlehtone » 2020/03/26 22:04:30

luiscssz wrote:
2020/03/26 20:36:16
but I am not sure if the release would be updated autmatically from 7.2 to 8. Do you have any experience with?
There is no supported way to upgrade CentOS from one major version to another. No 7->8. That you don't have to worry.

CentOS 7 is currently based on 7.7. Running "yum update" would go all the way to that. Are you sure that 7.7 is not an option? What in the 7.6->7.7 update does break the application?

CentOS does have an archive of old, unsupported releases. The Vault. Behind slow connection.
The "centos-release" package does define the (disabled) Vault repositories. Alas, you do need the 7.7's centos-release to get 7.6's vault's definition.

You have multiple nodes. Perhaps you can reserve one for experimenting. Make a backup of it first, should things go wrong.

I would bet that the application does not depend on yum. Therefore, I would update that first:

Code: Select all

sudo yum update yum\*
sudo yum update centos-release
sudo yum clean all
Now you should see the repositories:

Code: Select all

sudo yum repolist all
Then, you want to hide the 7.7's packages:

Code: Select all

sudo yum-config-manager --disable base updates extras
And you want to see the 7.6 content:

Code: Select all

sudo yum-config-manager --enable whatever_the_repo_names_are
Now "yum update" should do what you desire. Yum does show what it would do and does ask whether to proceed, so you can still abort.

You really should have kept your system up to date. The 7.2 ceased to receive any updates before December 2016. Major security holes have been found since then.

luiscssz
Posts: 7
Joined: 2020/03/26 18:42:04

Re: how to update from centos 7.2 to 7.6 in a cluster

Post by luiscssz » 2020/03/27 09:03:38

CentOS 7.7 is unfortunately not an option. As you recommend me, I would prefer to update to the last release, but as we simulate in another bigger cluster, in which centos 7.6 is installed, we have to replicate the same conditions in our cluster to avoid issues with different OS, softwares, etc...

About your ideas (should I open independent threads?):
  • 1.- How can I reserve 1 node for experimenting? I thought the update takes place in all nodes
  • 2.- How can I make a back-up of a node?
  • 3.- How can I specify the node where I install a software
What do you mean "the application does not depends on yum"? If you means ANSYS 19R2, you are right. This is the second big task for me: install ANSYS 19R2 in a cluster. Perhaps I should open a new thread with this second task. Correct me if I am wrong but there are 2 options for installing a software in a cluster:
  • 1.- install the software in the master and export with NFS to the other nodes (no idea how I can do that)
  • 2.- install the images in all nodes (no idea how I can do that)
You are complete right about the updates. A former collegue of my department has installed the cluster in 2016. In 2017 changed to other company and nobody needed/updated the cluster until now.

User avatar
jlehtone
Posts: 4531
Joined: 2007/12/11 08:17:33
Location: Finland

Re: how to update from centos 7.2 to 7.6 in a cluster

Post by jlehtone » 2020/03/27 12:41:31

luiscssz wrote:
2020/03/27 09:03:38
we simulate in another bigger cluster, in which centos 7.6 is installed, we have to replicate the same conditions in our cluster to avoid issues with different OS, softwares, etc...
Sounds like the requirement for 7.6 is not a fact set by Ansys, but a concern about numerical reproducibility. Fair enough.
luiscssz wrote:
2020/03/27 09:03:38
This is the second big task for me: install ANSYS 19R2 in a cluster. Perhaps I should open a new thread with this second task.
Support and instructions of a commercial product should come from the vendor.
luiscssz wrote:
2020/03/27 09:03:38
Correct me if I am wrong but there are 2 options for installing a software in a cluster:
  • 1.- install the software in the master and export with NFS to the other nodes (no idea how I can do that)
  • 2.- install the images in all nodes (no idea how I can do that)
Yes. Depends on how software is distributed.

The 'yum' manages "RPM-packages". RPM-packages contain enough metadata that they can be installed, updated and removed in controlled fashion. They are installed locally to each node.

Unmanaged content is better to put outside of system directories. One can still place it into each node, but if multiple machines can access the files simultaneously and network access is not a bottleneck, then common shared volume makes sense.


There are many ways to install and maintain a cluster. There are more than one type of clusters.

A cluster can have a scheduler that allocates resources for "jobs". Jobs are "submitted" and will execute when resources are available. The scheduler can be told that a specific node (a resource) is "down", "unavailable", "reserved for maintenance", etc. We do not know what possibilities your "cluster" offers.

The 'yum' does not know anything about clusters. It runs on "this machine". The solution is to run the same yum command on each node.


Do you know how the "another bigger cluster" is set up / maintained?


A backup is a copy from which a state can be restored from. That involves knowing what is essential for the state. Or perhaps just create a copy of the raw block device(s). Even then, one has to know what one has and one has to have a place to save the copy to.


Overall, I recommend getting an expert on-site.

Post Reply