Re: [ceph-users] Ubuntu upgrade Zesty => Aardvark, Implications for Ceph?

2017-11-21 Thread Ken Dreyer
As a tangent, it's a problem for download.ceph.com packages that
"xenial" happens to sort alphabetically after "bionic", because
do-release-upgrade will consider (for example) "ceph_12.2.1-1xenial"
to be newer than "ceph_12.2.1-1bionic". I think we need to switch to
using "ubuntu16.04", "ubuntu18.04" suffixes instead.

- Ken

On Mon, Nov 13, 2017 at 4:41 AM, Ranjan Ghosh  wrote:
> Hi everyone,
>
> In January, support for Ubuntu Zesty will run out and we're planning to
> upgrade our servers to Aardvark. We have a two-node-cluster (and one
> additional monitoring-only server) and we're using the packages that come
> with the distro. We have mounted CephFS on the same server with the kernel
> client in FSTab. AFAIK, Aardvark includes Ceph 12.0. What would happen if we
> used the usual "do-release-upgrade" to upgrade the servers one-by-one? I
> assume the procedure described here
> "http://ceph.com/releases/v12-2-0-luminous-released/; (section "Upgrade from
> Jewel or Kraken") probably won't work for us, because "do-release-upgrade"
> will upgrade all packages (including the ceph ones) at once and then reboots
> the machine. So we cannot really upgrade only the monitoring nodes. And I'd
> rather avoid switching to PPAs beforehand. So, what are the real
> consequences if we upgrade all servers one-by-one with "do-release-upgrade"
> and then reboot all the nodes? Is it only the downtime why this isnt
> recommended or do we lose data? Any other recommendations on how to tackle
> this?
>
> Thank you / BR
>
> Ranjan
>
>
>
>
> ___
> ceph-users mailing list
> ceph-users@lists.ceph.com
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com


Re: [ceph-users] Ubuntu upgrade Zesty => Aardvark, Implications for Ceph?

2017-11-19 Thread Gregory Farnum
On Mon, Nov 13, 2017 at 10:42 PM Ranjan Ghosh  wrote:

> Hi everyone,
>
> In January, support for Ubuntu Zesty will run out and we're planning to
> upgrade our servers to Aardvark. We have a two-node-cluster (and one
> additional monitoring-only server) and we're using the packages that come
> with the distro. We have mounted CephFS on the same server with the kernel
> client in FSTab. AFAIK, Aardvark includes Ceph 12.0. What would happen if
> we used the usual "do-release-upgrade" to upgrade the servers one-by-one?
> I assume the procedure described here
> "http://ceph.com/releases/v12-2-0-luminous-released/;
>  (section "Upgrade
> from Jewel or Kraken") probably won't work for us, because
> "do-release-upgrade" will upgrade all packages (including the ceph ones) at
> once and then reboots the machine. So we cannot really upgrade only the
> monitoring nodes. And I'd rather avoid switching to PPAs beforehand. So,
> what are the real consequences if we upgrade all servers one-by-one with
> "do-release-upgrade" and then reboot all the nodes? Is it only the downtime
> why this isnt recommended or do we lose data? Any other recommendations on
> how to tackle this?
>
It's just the downtime that prevents people doing stuff like this. If
that's not a concern for you, it won't hurt your data, although you may
need to poke at the services a bit to persuade them all to get going again.

Do keep in mind that while it should work, I'm not aware of anybody testing
this.
-Greg


> Thank you / BR
>
> Ranjan
>
>
>
> ___
> ceph-users mailing list
> ceph-users@lists.ceph.com
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com


[ceph-users] Ubuntu upgrade Zesty => Aardvark, Implications for Ceph?

2017-11-13 Thread Ranjan Ghosh

Hi everyone,

In January, support for Ubuntu Zesty will run out and we're planning to 
upgrade our servers to Aardvark. We have a two-node-cluster (and one 
additional monitoring-only server) and we're using the packages that 
come with the distro. We have mounted CephFS on the same server with the 
kernel client in FSTab. AFAIK, Aardvark includes Ceph 12.0. What would 
happen if we used the usual "do-release-upgrade" to upgrade the servers 
one-by-one? I assume the procedure described here 
"http://ceph.com/releases/v12-2-0-luminous-released/; (section "Upgrade 
from Jewel or Kraken") probably won't work for us, because 
"do-release-upgrade" will upgrade all packages (including the ceph ones) 
at once and then reboots the machine. So we cannot really upgrade only 
the monitoring nodes. And I'd rather avoid switching to PPAs beforehand. 
So, what are the real consequences if we upgrade all servers one-by-one 
with "do-release-upgrade" and then reboot all the nodes? Is it only the 
downtime why this isnt recommended or do we lose data? Any other 
recommendations on how to tackle this?


Thank you / BR

Ranjan



___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com