Strahil
2019-Sep-24 08:18 UTC
[Gluster-users] Where does Gluster capture the hostnames from?
Hi Tom,
Your test with dig is wrong - dig and nslookup always query a DNS server.
Maybe 'host' command can show it.
I'm using /etc/hosts only for critical systems - like any kind of clusters
as based on my experience - you never know when your DNS will be unreachable (in
my case it was a network issue, not a DNS actually).
Of course , a local (same subnet) DNS caching/forwarding server can help and
provide offloading for the DNS servers.
Also, I prefer to set all cluster memebers in ntp/chrony's config as peers -
because even if your time servers are down , the cluster will stay in sync which
is very important.
Have you tried setting an entry like this:
IP FQDN hostname
Then 'hostname' will show something like 'mdskvm-p01' instead of
FQDN.
Best Regards,
Strahil Nikolov
On Sep 24, 2019 04:31, TomK wrote: > > Hey All, > > My hosts below:
> > [root at mdskvm-p01 ~]# cat /etc/hosts > 127.0.0.1?? localhost
localhost.localdomain localhost4 > localhost4.localdomain4 > ::1????????
localhost localhost.localdomain localhost6 > localhost6.localdomain6 >
[root at mdskvm-p01 ~]# hostname > mdskvm-p01.nix.mds.xyz > [root at
mdskvm-p01 ~]# hostname -f > mdskvm-p01.nix.mds.xyz > [root at mdskvm-p01
~]# > > [root at mdskvm-p02 ~]# cat /etc/hosts > 127.0.0.1?? localhost
localhost.localdomain localhost4 > localhost4.localdomain4 > ::1????????
localhost localhost.localdomain localhost6 > localhost6.localdomain6 >
[root at mdskvm-p02 ~]# hostname > mdskvm-p02.nix.mds.xyz > [root at
mdskvm-p02 ~]# hostname -f > mdskvm-p02.nix.mds.xyz > [root at mdskvm-p02
~]# > > My take on the /etc/hosts file discussion: > > 1) If
hostname / hostname -f returns any valid values, the software > should
capture it. > > 2) There is no benefit or need to use /etc/hosts in a
small setup. > Larger setups resolving hosts against an enterprise DNS behind
many > switches could be a problem.? Managing our /etc/hosts files using >
Ansible helped to reduce some of these problems esp since lookups are >
logged against the connection tracking tables, that can get full, > network
response time could vary etc.? ("Semi static" I guess might >
describe this approach best?)? These are populated, if changes are > needed,
via an initial DNS lookup once a day. Invariably, managing > /etc/hosts is
time consuming and messy. > > 3) Running a good DNS cluster, something
like a two node IPA cluster > that I run for a small setup, prevents such
outages.? This particularly > when also placing a VIP across the nodes and
locating cluster nodes > across different hardware and locations. > >
4) Point 2) should be no reason why an application cannot obtain or > resolve
proper DNS entries in 1). > > Having said that, decided to check if
there's any benefit to having > entries in /etc/hosts: > > [root at
mdskvm-p01 ~]# time $(dig? mdskvm-p01.nix.mds.xyz >/dev/null) > >
real??? 0m0.092s > user??? 0m0.087s > sys???? 0m0.005s > [root at
mdskvm-p01 ~]# time $(dig? mdskvm-p02.nix.mds.xyz >/dev/null) > >
real??? 0m0.092s > user??? 0m0.084s > sys???? 0m0.008s > [root at
mdskvm-p01 ~]# cat /etc/hosts > 127.0.0.1?? localhost localhost.localdomain
localhost4 > localhost4.localdomain4 > ::1???????? localhost
localhost.localdomain localhost6 > localhost6.localdomain6 >
192.168.0.60??? mdskvm-p01.nix.mds.xyz? mdskvm-p01 > 192.168.0.39???
mdskvm-p02.nix.mds.xyz? mdskvm-p02 > [root at mdskvm-p01 ~]# vi /etc/hosts
> [root at mdskvm-p01 ~]# time $(dig? mdskvm-p01.nix.mds.xyz >/dev/null)
> > real??? 0m0.093s > user??? 0m0.082s > sys???? 0m0.010s >
[root at mdskvm-p01 ~]# time $(dig? mdskvm-p02.nix.mds.xyz >/dev/null) >
> real??? 0m0.093s > user??? 0m0.085s > sys???? 0m0.007s > [root at
mdskvm-p01 ~]# time $(dig? mdskvm-p01.nix.mds.xyz >/dev/null) > >
real??? 0m0.094s > user??? 0m0.084s > sys???? 0m0.010s > [root at
mdskvm-p01 ~]# time $(dig? mdskvm-p02.nix.mds.xyz >/dev/null) > >
real??? 0m0.092s > user??? 0m0.081s > sys???? 0m0.011s > [root at
mdskvm-p01 ~]# cat /etc/hosts > 127.0.0.1?? localhost localhost.localdomain
localhost4 > localhost4.localdomain4 > ::1???????? localhost
localhost.localdomain localhost6 > localhost6.localdomain6 > [root at
mdskvm-p01 ~]# > > So with /etc/hosts file entries present makes little
difference in small > setup when governed by /etc/nsswitch.conf . > >
Having entries in /etc/hosts, doesn't affect how gluster displays the >
entries when calling gluster volume status . > > Cheers, > TK > >
On 9/23/2019 11:36 AM, Joe Julian wrote: > > Perhaps I misread the intent,
I apologize if I did. I read "static > > entries" as "ip
addresses" which I've seen suggested (from my > > perspective)
far too often. /etc/hosts is a valid solution that can > > still adapt if
the network needs to evolve. > > > > On 9/23/19 8:29 AM, ROUVRAIS
Cedric wrote: > >> Hello, > >> > >> I gue
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20190924/0f3aefcf/attachment.html>