OCFS For Windows

Introduction

Usually “single-hosted” filesystems like NTFS or FAT assume they are the only one who accesses a volume after they mounted it. With this assumption they do not have to take care of other hosts doing modifications to data and metadata on the same filesystem. Using disks with these filesystems on more than one host would corrupt them pretty fast because neither host would take care of what the others do. OCFS provides mechanisms to support concurrent access to shared volumes from different hosts. This chapter also tries to provide some background information about OCFS probably not mentioned in other documents.

The procedure of mounting and volume recognition

Windows usually loads filesystem drivers during boot time. This does not mean that a volume is being mounted immediately. In fact mounting is usually delayed until a program or the OS itself accesses a volume the first time. When this happens Windows sends a message to each filesystem driver telling it to mount a specific volume. The filesystem driver in turn accesses the volume and tries to recognize the filesystem. If the filesystem is recognized it completes that mount request and further IO requests are send to that filesystem driver.Mounting can be (and usually is) even more complex. Usually a filesystem driver does not only do the filesystem recognition before it responds with “Success” to the OS. It may also check if the volume is “dirty” (which means the system was not shut down properly) or if the volume was just disabled because the administrator is doing maintenance work.

Major differences to “well-known” filesystems like NTFS or FAT

There are some differences between NTFS and OCFS one should be aware of.
Security
Unlike NTFS, OCFS has no build in support for ACLs. Neither on directory nor on file level. There is no directory or file security at all. The only security mechanism OCFS provides is that only users that belong to the “Administrators” group can access OCFS volumes at all.
Predictable “Hangs”
There are a few scenarios where one or more cluster nodes might appear to hang. It depends on the reason how long this hang will last.

A crash of one node might cause the other nodes to hang for up to 7 minutes. This is because the other nodes have to be sure the “dead node” is really dead and the locks it owned can never be acquired.
A regular shutdown or restart of one node will stop the other nodes for less than a minute.

Mountpoints & Driveletters

OCFS only supports one mountpoint per volume. It can either be a driveletter or linked from an NTFS partition. The mountpoint is assigned by Disk Administrator and stored on each OCFS volume during format. It is not sufficient to change a mountpoint in Disk Administrator only. One will have to change it using ocfsutil. Although the better approach is not to change mountpoints once the volume was formatted. This may create an inconsistent view onto the volumes from different nodes. NTFS links to OCFS volumes should be preferred over driveletters.