Forcibly Unmounting NFS Filesystems

forcibly unmounting nfs filesystems

preventative steps

preventative

steps

As discussed earlier in this paper, there are many reasons why an NFS filesystem may at

some point transition to a state where it cannot be unmounted; such as when processes

are holding resources open on the filesystem and cannot be killed. There are also times

when it appears no processes are holding any NFS filesystem resources and yet the

filesystem is still considered “busy.” We saw in the example beginning on page 7 how a

process can have active buffer cache memory resources pending against an NFS

filesystem and how this process may not show up in an fuser(1M) process list even though

it continues to hold these buffer cache pages. There have even been customer reported

cases where no processes are keeping NFS files open or holding buffer cache resources

against an NFS filesystem, yet the filesystem is still considered “busy.”

Whatever the cause, the result is the same – an NFS filesystem that cannot be unmounted

when desired. While some of these “busy” NFS filesystem scenarios are unavoidable,

such as server hardware or software failures, there are precautions you can take when

configuring your NFS clients and servers to avoid many of the other common situations

that cause NFS filesystems to be incorrectly considered “busy.” Even in the case of

server hardware or software failures, there are steps you can take to protect your NFS

clients from prolonged server outages.

install the latest

fuser(1M), NFS, and

virtual memory

patches

Periodically, NFS filesystems may remain in a “busy” state even though no processes that

are accessing this filesystem can be identified by fuser(1M). This situation could be

caused by any number of reasons – processes holding valid buffer cache resources but

do not have files open in the NFS filesystem,

a defect in the fuser(1M) code that hinders

its ability to report processes correctly, or defects in the NFS or Virtual Memory

subsystems that cause NFS filesystem data structures to become inaccurate, causing the

kernel to incorrectly consider a filesystem “busy” when it is really idle.

HP has released several patches to address these problems on HP-UX 11.0 and 11i

systems. The patches listed below in Table 1 were current as of the time of this writing;

however these patches may have since been superseded. In addition, several of the

patches listed in Table 1 are dependent on other patches for proper operation. Before

installing any of these patches on your HP-UX NFS clients and servers, check with HP

Support to obtain a current list of patches for your specific operating system, or use the

patch tools available at HP’s IT Resource Center Web site: http://itrc.hp.com

Table 1 - Recommended Patches to Avoid Known NFS umount(1M) Problems

Patch Name Supported OS Patch Description

PHCO_21901 11.0 fuser(1M) cumulative

PHNE_28567 11.0 ONC/NFS General Release/Performance

PHNE_28568 11i ONC/NFS General Release/Performance

PHKL_27266 11i iCOD, RTSCHED, (u)mount, final close, NFS umount

Refer to the example shown on pages 7 through 10.