On Thu, 27 Aug 2020 at 13:21, Anthony D'Atri <anthony.datri(a)gmail.com>
wrote:
Looking for a bit of guidance / approach to upgrading from Nautilus to
Octopus considering CentOS and Ceph-Ansible.
We're presently running a Nautilus cluster (all nodes / daemons 14.2.11
as
of this post).
- There are 4 monitor-hosts with mon, mgr, and dashboard functions
consolidated;
You want an odd number of mons. Add or remove one.
Agreed. Odd number is the target end state.
- 4 RGW hosts
- 4 ODS costs, with 10 OSDs each. This is planned to scale to 7 nodes
with additional OSDs and capacity (considering to do this as part of
upgrade process)
Don’t tempt fate. One thing at a time. Not three.
Never said I was doing all three. One at a time as per suggested proc.
We would be upgrading MGRs, MONs in one go given they are collocated on the
same nodes.
- Currently using ceph-ansible (however it's
a process to maintain
scripts
/ configs between playbook versions - although a
great framework, not
ideal
in our case;
^ Kefu ;)
?? Not sure I follow. Our question is around Ceph Orchestrator vs
Ansible. The idea of having something managed by the Ceph project vs. a
bolt-on. There are valid arguments for both. My comments were not
intended to offend. Our objective is to reduce complexity / moving parts
in managing ceph as a whole. Given the project has native orchestrator it
would be preferred to leverage / transition into that (for our deployment).
Octopus support on CentOS 7 is limited due to
python dependencies, as a
result we want to move to CentOS 8 or Ubuntu 20.04.
Do you have a compelling reason to go to Octopus today?
Is there a compelling reason not to proceed? Is it not the next stable
release? 4 updates since release so far. Specifically, I'm after object
lock and other performance efficiencies.
The other outlier is CentOS native Kernel
support for LSI2008 (eg.
9211) HBAs which some of our
OSD nodes use.
How is this a factor, do newer kernels drop support for that old HBA?
It's a RHEL / CentOS thing. Mainline and Ubuntu kernels support is just
fine. It's a mature HBA :) extensively deployed and used in scale out
storage clusters.
Here's an
upgrade path scenario that is being considered. At a
high-level:
I suggest that if you are set on doing this, you do one step at a time and
don’t try to get fancy. Especially since you only have one cluster.
Thats the intent. I'm looking for validation / experiences and others
from their upgrades.
I believe there are Nautlius packages available for CentOS 8 now, so
perhaps:
* Update each node — serially — to CentOS 8 + new Ceph packages
* Update to Octopus via the documented method
* Add your new nodes
Noted. It's a valid scenario as well.