[Deepsea-users] [ses-users] Deploying on 3 networks with DeepSea
jfajerski at suse.com
Mon Nov 6 14:59:23 MST 2017
On Mon, Nov 06, 2017 at 02:34:17PM -0700, Ricardo Dias wrote:
>> On 6 Nov 2017, at 14:17, Jan Fajerski <jfajerski at suse.com> wrote:
>> On Mon, Nov 06, 2017 at 02:00:38PM -0700, Ricardo Dias wrote:
>>>> On 6 Nov 2017, at 13:55, Jan Fajerski <jfajerski at suse.com> wrote:
>>>> Hi list,
>>>> does the following scenario ring any alarm bells with anyone?
>>>> The cluster has a salt-master, MONs and OSDs. All nodes have 3 networks: a management network and public- and cluster-network. All nodes have a hostname that resolves via DNS to the ip on the management network. The ips on cluster- and public-network don't correspond to DNS names.
>>>> The MONs derive their IDs from the short hostname, which would resolve to the ip addresses on the management network if one would attempt that.
>>>> The cluster deployed and operates fine. Is there any conceivable disadvantage with this setup? For example:
>>>> Should there be dns names for the public-network ip addresses?
>>>> Does DeepSea have any issue (or will have issues in the future) with running on the management network (the salt master has access to the public-network)?
>>> The only thing I’m seeing is that the public network might not be used at all because the MONs will bind to the management network, and the OSDs will connect to the MONs using the management network.
>>> If the management network has a lower bandwidth than the public network, then it might be a problem.
>> ceph.conf contains the mon_host settings with the appropriate ip addresses on the public network. So I'd assume the MONs use the correct network. Iiuc the mon_initial_members list contains only a list of MON ids and has nothing to do with networking.
>Was ceph.conf generated by DeepSea or it was tweaked manually?
>I’m asking this because with the network setup you described, I was expecting DeepSea to fill the public cluster setting in ceph.conf with the management network.
ceph.conf was generated by DeepSea but the public and cluster network was set to
the correct network after stage 1 in the pillar. I.e. DeepSea generated the
correct ceph.conf after being told what networks to use.
>>>> Imho this setup should work fine, but I might be missing something.
>>>> Jan Fajerski
>>>> Engineer Enterprise Storage
>>>> SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton,
>>>> HRB 21284 (AG Nürnberg)
>>>> Deepsea-users mailing list
>>>> Deepsea-users at lists.suse.com
>> Jan Fajerski
>> Engineer Enterprise Storage
>> SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton,
>> HRB 21284 (AG Nürnberg)
Engineer Enterprise Storage
SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton,
HRB 21284 (AG Nürnberg)
More information about the Deepsea-users