Hi List, I suppose this isn't strictly CentOS but I'm talking about some 5.5 servers so I thought someone else here may have had the same issues; We power on servers at night via IPMI and shut them down the same way for automation. The problem I'm facing is that the servers are shutdown in about 2 seconds. IPMI provides remote access to the power features as we know so I don't think the OS has a chance to do a 'graceful' shutdown. One server in particular is told to shutdown at say 07:00am, the last scheduled task on there should finish around 06:30 so there is a half hour window for over run. It now has a corrupt XFS file system which I am repairing as we speak. I am wondering if this could be a possible cause of these problems? Perhaps if that last tasks over runs a couple of times and it gets powered off in the middle of the tasks? (These are backup servers rsyncing from other servers) Does anyone here have problems with this? -- James. http://www.jamesbensley.co.cc/ There are 10 kinds of people in the world; Those who understand Vigesimal, and J others...?
On Wed, 2011-04-06 at 11:06 +0100, James Bensley wrote:> Hi List, > We power on servers at night via IPMI and shut them down the same way > for automation. The problem I'm facing is that the servers are > shutdown in about 2 seconds. IPMI provides remote access to the power > features as we know so I don't think the OS has a chance to do a > 'graceful' shutdown.??? Why are you doing this? Don't do this. Just do a "shutdown" at the OS level - which will power off the server on 99.44% of hardware. Then power-on using IPMI.> One server in particular is told to shutdown at say 07:00am, the last > scheduled task on there should finish around 06:30 so there is a half > hour window for over run. It now has a corrupt XFS file system which I > am repairing as we speak. I am wondering if this could be a possible > cause of these problems?Yes. What you are doing is *evil*.> Perhaps if that last tasks over runs a couple > of times and it gets powered off in the middle of the tasks? (These > are backup servers rsyncing from other servers)Not to mention that your task possibly didn't finish.> Does anyone here have problems with this?Yes. What you are doing is bad and wrong. "shutdown -h now" will give you an immediate shutdown and power-off the box. You can also schedule shutdown via shutdown, see "man shutdown".
On Apr 6, 2011, at 6:06 AM, James Bensley <jwbensley at gmail.com> wrote:> Hi List, > > I suppose this isn't strictly CentOS but I'm talking about some 5.5 > servers so I thought someone else here may have had the same issues; > > We power on servers at night via IPMI and shut them down the same way > for automation. The problem I'm facing is that the servers are > shutdown in about 2 seconds. IPMI provides remote access to the power > features as we know so I don't think the OS has a chance to do a > 'graceful' shutdown. > > One server in particular is told to shutdown at say 07:00am, the last > scheduled task on there should finish around 06:30 so there is a half > hour window for over run. It now has a corrupt XFS file system which I > am repairing as we speak. I am wondering if this could be a possible > cause of these problems? Perhaps if that last tasks over runs a couple > of times and it gets powered off in the middle of the tasks? (These > are backup servers rsyncing from other servers) > > Does anyone here have problems with this?File systems are memory based objects that (loosely) don't exist on disk until shutdown. You yank out the power, who knows what you'll get, and it doesn't matter when the last job ran because the OS can be running house keeping functions on the file system in the background. -Ross