On Wed, 13 Jan 2010, Jeffrey Altman wrote:

On 1/13/2010 3:20 PM, Buhrmaster, Gary wrote:

For the time being at least, set the RxMaxMTU value back to 1260 octets.
 This will result in a performance penalty but will avoid this data
corruption.   Hopefully one night soon I will wake up in the middle of
the night understanding where and how this data corruption is taking place.

You mention SEP 11.0.4.  Does the problem still exist in release update 5?

Richard said in his e-mail that the problem could not be recreated with
11.0.5.

To followup with respect to SEP 11.0.5, we have never seen file corruption, but we have seen temporary network lockup ( a known issue with older SEP versions too) and general performance degradation with some combinations of parameters.

Interface-MTU RxMaxMTU Client-threads  Network-bandwith-utilization

not set       0        4               25-45%, no lockups (default settings)
not set       0        4               30-50%, no lockup, exempt AFScache
                                               from SEP scanning
1300          0        4               0-50%, network lockups
1300          0        1               0-50%, network lockups
1300          1260     1               25-40%, no lockups

This is based on tests with just two machines, so should not be taken as definitive, but it seems that SEP 11.0.5 does not cause problems on a machine with the default MTU configuration.


I ask because of the following "interesting" fixes in RU5:

http://service1.symantec.com/SUPPORT/ent-security.nsf/docid/2007121216360648


With NICs that use a TCP offload engine, Symantec Endpoint Protection with 
Network Threat Protection enabled causes networking problems, such as 
connection failures and performance degradation
Fix ID: 1389258
Symptom: Teefer2 causes packet loss with TCP/UDP checksum offload by not 
preserving checksum data.
Solution: Teefer2 corrected to preserve checksum data.


DNS resolution fails while connected via Microsoft VPN
Fix ID: 1442277
Symptom: Teefer2 causes packet loss with TCP/UDP checksum offload by not 
preserving checksum data.
Solution: Teefer2 corrected to preserve checksum data.

--
Richard Brittain,  Research Computing Group,
                   Kiewit Computing Services, 6224 Baker/Berry Library
                   Dartmouth College, Hanover NH 03755
[email protected] 6-2085
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info

Reply via email to