Reload ext3 file system
 

Hi guys,

I have a problem with sharing a file system. We have a SAN storage, directly attached to some servers. I would like to share the data among the hosts. Now we use GFS for this, but we wants to outsource the servers, and the outsourcing company couldn't handle any shared file system (GFS, VCFS, etc..), so I have to find out an alternative.

My idea is the following, because we need the data only for read, I would like to use ext3, and mount it only with read only. Of course one node will have rw access, and it will feed the storage with data.

I set up the system, I can mount the storage on every nodes, but when I write data to the storage, it is not visible on the other hosts. I know it is normal, because ext3 was not designed for this, but if I do umount and mount on the ro hosts, files will be available. I suppose behind that is that with mounting again the volume the kernel reads again the filesystem, and new files will be visible.

My question would be, are there any ways to reread the ext3 filesystem without unmounting and mounting? With this my problem is, if there is an open file from this volume, umount will not work.

I tried "sync" and "mount -o ro,remount" as well, but it does not read again the file system.

Thanks for any help in advance!

Cheers,

Chris

Do not go there.
Go see whoever is responsible for the decision to outsource, and ask them how much your data integrity (business ???) is worth.

Then tell them to get their head out of their @rse.

Yep, I know, you are right, but contract is already done. We have to live with this.

Any idea how to refresh the inode table?

Nope.
And it gets worse. Any system that finds the filesystem marked as dirty at mount time will attempt to run the journal. Mounting as read only doesn't affect this behaviour, and you can't stop it. This was a design decision by Ted I later found out (I was only doing lab testing, so it didn't matter to me).
Can you trust your provider to ensure that the filesystem mounted r/w is always the last unmounted, and the first mounted ?. What about crashes ...

I know you're in a bind. but you really don't want to be involved in this. Seems you had a good system setup, and you've now got to make a cockup work. Make sure you document all you try, and get it signed off by your manager. I reckon I know who'll wear the blame otherwise.

Really thank you that you highlighted this journal recovery case, I didn't know how it is working. We just discussed it, we can live with because we will write only temporary files to this disk, nothing important, we just want to share them, it is not a big problem if we loose the last X files because of a file system corruption.

I will progress with 5 seconds umount and mount.

Cheers,

Chris