597913 - stop mounting NFS on slaves/masters

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[[e-mail] (:bhearsum)]

We mounted NFS For serving puppet files for quite awhile. At this time, I don't think we serve anything on it anymore, and we should stop mounting it.

We still use it to share files between production-puppet/staging-puppet, so I'm not sure what to do there.

Comment 1

[matthew zeier [:mrz]]

Possible dup of bug 597912.  

I'm asked for this several times - this method just won't scale.  You can't expect hundreds of hosts to reliably connect to a remote NFS host.  IT doesn't use NFS for file distribution.

Comment 2

[Nick Thomas [:nthomas] (UTC+12)]

Is bug 597912 intentionally closed ?

Comment 3

[matthew zeier [:mrz]]

No.  I cleared the flags on it.

Comment 4

[[e-mail] (:bhearsum)]

I'll be doing this, still don't have an ETA.

Comment 5

[[e-mail] (:bhearsum)]

Attachment: die nfs die

This should remove all of our mounting of NFS. As far as I can tell, the only place that still mounts it is 32-bit linux build machines, as I did some removal of NFS in bug 552058. So, this patch:
- drops the unused file virtualenv.pp which references /N
- drops the puppet-files-* files/classes and all references to them
- ensures that /N is absent using the mount {} type for 32 and 64-bit linux build machines
- ensure /etc/fstab is absent on all macs

Additionally, there's a couple of file updates that need to be done on the masters:
- remove etc/fstab from all mac filestores
- update etc/auto_master on build macs to remove /N references on all mac filestores

Need to give this a quick run through staging still.

Comment 6

[Rail Aliiev [:rail]]

Comment on attachment 482596 [details] [diff] [review]
die nfs die

looks good

Comment 7

[[e-mail] (:bhearsum)]

Comment on attachment 482596 [details] [diff] [review]
die nfs die

changeset:   237:6f5d7eda6e9d

Comment 8

[[e-mail] (:bhearsum)]

This is landing cleanly. I need to go around to all the ref machines and verify that they pick it up; we'll also need to wait a couple of days to make sure all the slaves have rebooted.

Comment 9

[[e-mail] (:bhearsum)]

Got the following ref machines:
- talos-r3-{fed,fed64,leopard}-ref
- bm-mini-build-ref
- moz2-darwin10-ref

Still to do:
- linux-ix-ref
- talos-r3-snow-ref

Purposely ignoring (because I highly doubt we'll be cloning new VMs):
- linux-ref-platform
- linux64-ref-platform

Comment 10

[[e-mail] (:bhearsum)]

(In reply to comment #9)
> Still to do:
> - linux-ix-ref
> - talos-r3-snow-ref

These are done. Filed a bug to get an idea of who is still mounting the share -- should only be buildbot masters and two puppet masters at this point. Want to make sure that there's no slaves mounting it still before we block it at the firewall level.

Comment 11

[[e-mail] (:bhearsum)]

On second thought, it's probably not important to block this at the firewall level.