-----Original Message----- From: Rita [mailto:rmorgan466@gmail.com] Sent: Tuesday, November 12, 2013 7:22 PM To: CentOS mailing list Subject: Re: [CentOS] NFS hard mount
Yes, unmount the share and error to the client app.
Sounds like you want to have a combination of mounting with the intr flag[1] and use autofs[2] so that the file system is not mounted ALL the time, but not so soft as to loose data all the time. I have seen soft loose data reliably, were in the same environment (and same equipment) hard,intr does not loose data. My experience is that intr is useful in situations where servers go down/have link issues frequently, and autofs with short timeouts (~15s < timeout <~60s) makes those issues even less _visible_. It will not get you an automatic unmount of the share and error to the client app in the case of server down, but it will /allow/ you to terminate the client and unmount as needed IIRC.
[1] nfs mounting page with thanks to zGreenfelder https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/... [2] autofs page https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/...
On Tue, Nov 12, 2013 at 6:25 PM, John R Pierce pierce@hogranch.com wrote:
On 11/12/2013 3:18 PM, Rita wrote:
Is there a tunable to set NFS hard mount time out? For instance, if the server becomes unavailable for extended amount of time, say 1 hour. I
would
like to timeout after 30 secs.
timeout and do what, unmount the share, and error to the client app thats trying to do a read(),write(),open(),stat(),etc ?
-- john r pierce 37N 122W somewhere on the middle of the left coast
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- --- Get your facts first, then you can distort them as you please.--
Even when this disclaimer is not here: I am not a contracting officer. I do not have authority to make or modify the terms of any contract.
Thanks for the detailed response.
My problem is we have many NFS servers and clients. The client mounts to the servers using /net/serverA/dirA. When server A is decommissioned all the clients hand on `df`. I was wondering if there is a clean way to unmount the server instead of rebooting clients (100+ clients). I am try doing a lazy unmount but that doesn't seem to fix the problem.
On Wed, Nov 13, 2013 at 12:58 PM, Denniston, Todd A CIV NAVSURFWARCENDIV Crane todd.denniston@navy.mil wrote:
-----Original Message----- From: Rita [mailto:rmorgan466@gmail.com] Sent: Tuesday, November 12, 2013 7:22 PM To: CentOS mailing list Subject: Re: [CentOS] NFS hard mount
Yes, unmount the share and error to the client app.
Sounds like you want to have a combination of mounting with the intr flag[1] and use autofs[2] so that the file system is not mounted ALL the time, but not so soft as to loose data all the time. I have seen soft loose data reliably, were in the same environment (and same equipment) hard,intr does not loose data. My experience is that intr is useful in situations where servers go down/have link issues frequently, and autofs with short timeouts (~15s < timeout <~60s) makes those issues even less _visible_. It will not get you an automatic unmount of the share and error to the client app in the case of server down, but it will /allow/ you to terminate the client and unmount as needed IIRC.
[1] nfs mounting page with thanks to zGreenfelder
https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/... [2] autofs page
https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/...
On Tue, Nov 12, 2013 at 6:25 PM, John R Pierce pierce@hogranch.com
wrote:
On 11/12/2013 3:18 PM, Rita wrote:
Is there a tunable to set NFS hard mount time out? For instance, if
the
server becomes unavailable for extended amount of time, say 1 hour. I
would
like to timeout after 30 secs.
timeout and do what, unmount the share, and error to the client app thats trying to do a read(),write(),open(),stat(),etc ?
-- john r pierce 37N 122W somewhere on the middle of the left coast
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- --- Get your facts first, then you can distort them as you please.--
Even when this disclaimer is not here: I am not a contracting officer. I do not have authority to make or modify the terms of any contract.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
From: Rita rmorgan466@gmail.com
My problem is we have many NFS servers and clients. The client mounts to the servers using /net/serverA/dirA. When server A is decommissioned all the clients hand on `df`. I was wondering if there is a clean way to unmount the server instead of rebooting clients (100+ clients). I am try doing a lazy unmount but that doesn't seem to fix the problem.
Why not ifup the decommissioned IP on another server/pc just long enough for the clients to umount? I think you do not even need to have a nfs server listing on it...
JD
On Thu, Nov 14, 2013 at 5:34 AM, Rita rmorgan466@gmail.com wrote:
Thanks for the detailed response.
My problem is we have many NFS servers and clients. The client mounts to the servers using /net/serverA/dirA. When server A is decommissioned all the clients hand on `df`. I was wondering if there is a clean way to unmount the server instead of rebooting clients (100+ clients). I am try doing a lazy unmount but that doesn't seem to fix the problem.
I'd suggest that the NFS servers should have showmount -a run before hand and the clients unmount/changed before the decom happens... but then I'm bit by the seemingly broken nature of showmount on centos (at least for me on my 6.4 machines, and assuming you run centos as your NFS servers).
I'd think you keep a comprehensive list of clients, perhaps running through them to check for mounts to soon to be decom'd servers?
On Thu, Nov 14, 2013 at 11:59 AM, zGreenfelder zgreenfelder@gmail.com wrote:
On Thu, Nov 14, 2013 at 5:34 AM, Rita rmorgan466@gmail.com wrote:
Thanks for the detailed response.
My problem is we have many NFS servers and clients. The client mounts to the servers using /net/serverA/dirA. When server A is decommissioned all the clients hand on `df`. I was wondering if there is a clean way to unmount the server instead of rebooting clients (100+ clients). I am try doing a lazy unmount but that doesn't seem to fix the problem.
I'd suggest that the NFS servers should have showmount -a run before hand and the clients unmount/changed before the decom happens... but then I'm bit by the seemingly broken nature of showmount on centos (at least for me on my 6.4 machines, and assuming you run centos as your NFS servers).
I'd think you keep a comprehensive list of clients, perhaps running through them to check for mounts to soon to be decom'd servers?
I've usually had trouble unmounting anything with open files or current directories across the mount point. If there aren't, using the automounter would probably have worked. Are there versions where umount -f actually works?
-
Denniston, Todd A CIV NAVSURFWARCENDIV Crane -
John Doe -
Les Mikesell -
Rita -
zGreenfelder