WAFS/WAAS File Locking

mlouis · ‎04-28-2008

I have run into an issue that i believe is file server related however i just want to run it by the board for clarification.

A remote site was recently converted from a standalone server site to one that is supported via WAFS/WAAS 4.0.11 installed on a 512 WAE w/ 2GB of Ram acting as wafs edge. The core is a 612 w/ 4GB of ram.

All user home folders reside across the WAN in the data center on a NetApp NAS SAN hosting the home folder natively (there is no windows server involved).

The issue is with file locking. Periodically, when a user opens a file on the server that is already been opened they are not getting any indication that the file is in use. Once they go to save the file they are told by the system that they cannot save the file using the same filename since it is being modified.

From reading the docs on WAFS/WAAS this looks like it is a problem with the OpLocks coming from the NetApp NAS and nothing to do with WAFS since WAFS/WAAS doesn't act on file locking requests, only passing them through to the file server.

Are my assumptions correct in this regard? Could WAFS be interferring some how with this communication and causing the file locking information being received from the NetApp NAS not to be correctly sent to the client?

Any way for me to debug this besides a packet capture? Any show commands that could indicate an issue? or logs?

TIA

Mike Louis

Zach Seils · ‎04-28-2008

Mike,

In general, WAFS will relegate locking to the origin file server. A packet capture would be best here, but you can also use the following CLI commands to get started:

Verify that the CIFS session is being handled by WAFS:

show cifs session list

Verify what type of oplock is present for the file:

show stat wafs expert "-server Rx -mbean CifsFileSystemDB -oper query"

This command will return a list of all files opened through the cache. The 'Optimizations level' indicates the type of oplock:

0 = None

1= Exclusive

2 = Batch

3 = Level 2

Let me know if you have any questions, or feel free to send over a packet capture.

Zach

mlouis · ‎05-02-2008

Thanks Zach,

I am working on getting a packet capture collected. I will let you know what i find.

mike

mlouis · ‎05-09-2008

Zach,

Neither of these commands worked for me in the CLI. I am running 4.0.11 rev of code. Are they available in that version?

Zach Seils · ‎05-13-2008

Mike,

These commands were added in the 4.0.13 release.

Zach

mlouis · ‎05-13-2008

Zach,

Part of the problem appears to be packet loss of the 4050 messages. We have created a class-map to match 4050 traffic and the connectivity directive in the WAE. Hopefully this will solve some of the intermittent file locking issues we are having.

However, i had some other questions after poking around in the /expert mode of the WAE.

1. When changing the WAFS connectivity bandwidth statement the CM states that the device Edge Component must be restarted. Is there a command that will allow me to do this without rebooting the box? Can i just remove the device from the Core - Edge group and then reassign it to restart the process?

2. Under the expert mode in 4.0.11 i noticed this value:

LockManager MBean Attributes:

Name Type Access Value

MaxGrantedLock int RO 10

MaxPendingLock int RO 0

What does MaxGrantedLock mean here? Will it not grant more than 10 locks at one time?

I have about 60 simultaneous users on file shares at any one time and they are intermittently not receiving the "file in use" statement from the server. Could this be part of the problem?

Thanks in advance

Mike

mlouis · ‎05-13-2008

Zach,

Part of the problem appears to be packet loss of the 4050 messages. We have created a class-map to match 4050 traffic and the connectivity directive in the WAE. Hopefully this will solve some of the intermittent file locking issues we are having.

However, i had some other questions after poking around in the /expert mode of the WAE.

1. When changing the WAFS connectivity bandwidth statement the CM states that the device Edge Component must be restarted. Is there a command that will allow me to do this without rebooting the box? Can i just remove the device from the Core - Edge group and then reassign it to restart the process?

2. Under the expert mode in 4.0.11 i noticed this value:

LockManager MBean Attributes:

Name Type Access Value

MaxGrantedLock int RO 10

MaxPendingLock int RO 0

What does MaxGrantedLock mean here? Will it not grant more than 10 locks at one time?

I have about 60 simultaneous users on file shares at any one time and they are intermittently not receiving the "file in use" statement from the server. Could this be part of the problem?

Thanks in advance

Mike

Zach Seils · ‎05-15-2008

Mike,

Both of these values are just counters of the number of active and pending locks.

Zach