Understanding namespace validation and correction

zFS provides namespace validation and correction in a shared file system environment. First, it is important to understand the concept of a namespace. zFS communicates between sysplex members using XCF protocols. The zFS XCF protocol exchanges information among members about zFS ownership and other attributes of zFS mounted file systems. This information, which is kept in the memory of each zFS member, is called the zFS namespace. If zFS members do not agree on the zFS owner of each file system, there might be problems that require a zFS restart or an IPL to recover.

zFS namespace validation is invoked in one of four ways:

When an administration command experiences an XCF message timeout
Automatically at zFS initialization
Automatically when zFS detects a problem that might be because of a namespace inconsistency
Explicitly using the MODIFY ZFS,NSVALIDATE operator command.

zFS namespace validation compares the information that is stored in each zFS member. If zFS validation detects an inconsistency, one or more messages can occur (for example, IOEZ00612I) and zFS attempts to correct the inconsistency, using one of the following actions:

Updating the inconsistent information
Automatically remounting a file system
Internally restarting zFS on one or more members.

The corrective action is disruptive and might cause one or more applications to receive I/O errors and display messages IOEZ00618E through IOEZ00637E. In addition, zFS might take SVC dumps when it detects a name inconsistency; therefore, do not issue the MODIFY ZFS,DUMP,ALL command.

Each zFS only keeps track of file systems that are locally mounted. z/OS® UNIX locally mounts file systems on systems where the mount was issued (or directed to through the SYSNAME parameter), and for sysplex-aware file systems, on other systems. z/OS UNIX keeps mount information that is hardened in the couple data set. In addition, zFS keeps track of zFS ownership by using cross system ENQ. The zFS owner of an aggregate always has an exclusive ENQ with a qname of SYSZIOEZ and an rname of IOEZLT.aggregatename. In this way, zFS hardens zFS ownership information in an independent repository. When an inconsistency is detected in the zFS namespace information between zFS members, this hardened information can be queried to determine how to automatically correct the inconsistency.

Tip: You can use the DISPLAY GRS,RES=(SYSZIOEZ,*) operator command to display zFS ENQs. For RNAME explanations and use, see the topic on Serialization summary in z/OS MVS Diagnosis: Reference.

Requirement: Because of the enhanced directory XCF protocol between zFS members, and because zFS always runs as sysplex=filesys, two additional steps must be performed on z/OS V1R12 to work with zFS on z/OS V1R13 and later releases:

You must apply toleration APAR OA39466.
zFS on V1R12 must run at sysplex=filesys to allow zFS on z/OS V1R13 and later releases to enter the shared file system environment.

For more information and migration instructions, see Applying required APARs for z/OS V2R1 and z/OS Migration.