[Beowulf] CentOS 7.x for cluster nodes ?

[Michael Di Domenico]

On Fri, Dec 30, 2016 at 1:12 AM, Andrew Mather <mathera at gmail.com> wrote:
>
> When v7 of the Redhat-based OS' appeared, the change to systemd in
> particular, seemed to attract a lot of hate, but since it's been out a
> while, there doesn't seem to be as much.

i think the shouting is less now mostly because there really isn't an
alternative.  devaun is out and running, i can't speak to uptake
though.  i believe if centos had done the same and split off a
separate distro without systemd in a similar manner to devuan the
landscape would look different.

> So, has anyone got recent war-stories, good experiences etc to share about
> v7 of CentOS specifically as the OS for cluster nodes.
>
> We don't have infiniband interconnects and don't use MPI, shared memory and
> the like.  All our jobs stay within the confines of the nodes and we have a
> variety of hardware configurations to accommodate different types of job
> (RAM, disk requirements etc)

given your fairly generic environment, i suspect you'll have little
difficultly going from v6 to v7.

i've been fighting v6 to v7 issues on three separate networks for the
last 2yrs.  some of which are systemd related but many more are just
the v6 to v7 jump and v7.2 to v7.3 was a new ball of wax.  though, i
have a varied and highly customized environment.  we've uncovered
several bugs in rhel for things like X, nfs/autofs, and some others.
many related to new hardware with half-baked driver support

one of the things i'll suggest, that we choose poorly on, was to keep
intact some of the v6 things in v7.  like iptables over firewalld,
keeping the nic/device names in the v6 way, removal of network
manager, and some other things like that.  we're starting to see the
ramifications now of those choices (i'll toot my own horn to say i saw
those issues wayback but was vetoed in the original decision)

unfortunately you can't predict what might break in the v6 to v7 jump.
we tried and still find things that trip us up, though it seems to be
declining.  it used to be weekly or daily that someone screamed in
agony down the hall at something stupid in v7, now it's probably
monthly