| 1 |
On Thursday 25 March 2004 03:02 pm, Sebastien Arnaud wrote: |
| 2 |
> Hi, |
| 3 |
> |
| 4 |
> I recently setup a few new Gentoo servers at a datacenter where we host |
| 5 |
> all our production environment. Today, for some reason we had to hard |
| 6 |
> reboot one of them, but it never came back up. A technician had to |
| 7 |
> drive down to the datacenter to see that some corruption had occurred |
| 8 |
> on the disk (reiserfs on the / mount point) and that it was waiting for |
| 9 |
> a root password or Control D to continue. |
| 10 |
> Anyway to change this behavior to make sure that the server always |
| 11 |
> boots and launches the SSH daemon to allow remote access to fix any |
| 12 |
> issues? |
| 13 |
|
| 14 |
No, because you cannot mount the / partition to load the enviroment and ssh |
| 15 |
and still do a filesystem check. Also, usually when you hit that point, it's |
| 16 |
not something that is going to boot without manual intervention. Usually, |
| 17 |
ctrl+d doesn't boot the machine. |
| 18 |
|
| 19 |
Part of the reason it does this is to give you a chance to repair a failed |
| 20 |
filesystem before it trys to mount it. You _could_ change your fstab options |
| 21 |
and tell it not to check on boot, but I wouldn't do. |
| 22 |
|
| 23 |
"The sixth field (0) is used by fsck (the filesystem check utility) to |
| 24 |
determine the order in which filesystems should be checked. |
| 25 |
If zero then fsck won't check the filesystem." |
| 26 |
|
| 27 |
> |
| 28 |
> On the same topic, does anybody know how to setup in Gentoo a console |
| 29 |
> access through one of our serial COM ports ? I would like to be able to |
| 30 |
> get access to the console of the Gentoo servers in case for some reason |
| 31 |
> the server does not start properly and does not launch the SSH daemon |
| 32 |
> for remote admin. |
| 33 |
|
| 34 |
The best equipment for total remote administration would be KVM over IP. |
| 35 |
Quite pricey though, sometimes it's better to just have a local tech or use |
| 36 |
the DC's remote hands, or go with a company that offers a bit more service to |
| 37 |
co-location. |
| 38 |
|
| 39 |
Honestly, out of the 100 or so servers I manage in our datacenter, I might run |
| 40 |
into that problem twice a year. We use ext3, but I can't see reiserfs being |
| 41 |
all that bad either. The worst I've ever seen was one co-lo client, it took |
| 42 |
several hours to repair his disk, but typically its a rare problem for us. |
| 43 |
|
| 44 |
|
| 45 |
Rob |
Archives