Informatica
NFS File Systems
Hello All,
As Informatica administrator, recommend admins to know information on file systems (Shared) on Informatica Servers, most of the
jobs read/write on share shared file systems and this will cause Latency in
jobs and performance is degraded.
Below are few inputs on NFS from Informatica,
please make sure to discuss with your Storage or Linux Admins for any
latency issues.
• NFSv2 has several
limitations in terms of locking, reliability and in ability to access 64-bit
file offsets. Also, industry has mostly migrated to NFSv3 and v4.
• NFSv3 has improved file
locking support over v2 but remains a stateless protocol dependent on a
separate lock manager – this can cause issues with recovery scenarios. Primary
concern is with single point of failure in the NFS server. Based on discussions
with Netapp & EMC this is resolved in most NAS appliances by integrating
the NFS server support into their proprietary file system. Open concern still
exists for standalone NFS server processes. Currently supported with Netapp
WAFL NAS and EMC Celerra NAS with some caveats. Please see supported shared
file system section.
• NFSv4’s new design seems
have resolved lingering issues with NFSv3’s locking and recovery and is the
preferred option for NFS-driven solutions.
• Critical - maintaining a
write lock on the master pmserver lck file used to arbitrate split brain
This indicates a possible latency
issue in using the NFSv4 file system between the time the session creating the
dynamic mapping completes and the one using it starts and needs to read it. A
solution would be to use the Timer task to insert a delay between the sessions
to allow the file system to complete before the second session tried to read the
file.
Thanks for sharing your honest opinion, you did a great job __
ReplyDelete