SRDF/A 'Partitioned' vs 'TransIdle'

Question

Can somebody explain the difference between these two states to me, please? I can't find it in manual or in Forums (or can't figure it out anyway).

We have been having line problems between primary and D/R site. When this happens the "big/active" SRDF groups go "Partitioned" and show Invalid Tracks, but the "small/low-activity" RDFGs go "TransIdle". As soon as the !@#$% Network Team fixes the line, the "TransIdle" RDFGs pick up with no problem, and the "Partitioned" groups go "Suspended" and have to be manually caught up (disable, set mode acp_disk, establish, set mode async, enable...). In NO case do I ever lose anything or have to do any "full establish".

So, what is happening here? In both cases all changes are stored on the Source Array. In both cases everything catches up (bigger RDFGs take longer, of course). So, practically, they both have the same characteristics. Except one restarts automatically, and one I have to restart manually.

sracharla · Answer

Hi stuart,

I cuold think of the following reason, the big SRDF group might be consuming high cache exceding the SRDF/A maximum cache utilization limits. And when that happens, the SRDF/A session drops and goes to partitioned state from transidle state.

Whereas the smaller SRDF group remains in transidle state all through beacuse its cache utilzation would be less within the limits.

Have a look at the cache utilzation of the SRDF groups during the link down time to confirm this.

Thanks,

Santhosh.

sracharla · Answer

you can look that in SPA. Navigate to SPA snapshot views and select the RDFA group. You can see the RDFA Write Pending usage and WP counts & DSE Counts graphs. You may want to look at the diagnostic views also if necessary.

Thanks

StuartA58 · Answer

How do I 'look at the cache utilization of the SRDF groups'? Is there a command for that?    Stuart

RobertDudley · Answer

From the help section of SMC:

Partitioned

The SYMAPI is currently unable to communicate through the corresponding RDF path to the remote Symmetrix. Partitioned may apply to devices within an SRDF group.

For example, if SYMAPI is unable to communicate to a remote Symmetrix via an SRDF group, devices in that RA group will be marked as being in the Partitioned state.

Transmit Idle Time — enables the data to be sent to cache if the link is idle. The system keeps retrying to send the data. This must be enabled on both local and remote sides.

For the lower usage groups they may not have the change rate to actually drop when you have a link outage. The larger group goes into partitioned state; basically saying I lost comm to the remote array. When the link is returned the devices that are partitioned go to suspended state as they are now suspended; the source array knows the status of the target but its not transmitting data.

ravi_paladugu · Answer

thats a good piece of article, just want to add couple more lines about tranmit idle - The SRDF/A session cannot push data in the transmit cycle across the link because the link is down.

GEMC · Answer

This may sound naive!   Are they two different RDFGs? If yes, have we enabled Transmit Idle (rdfa_transmit_idle) parameter on active RDFG? It works on RDFG group basis and it needs to be turned on both local and remote RDFGs.   Transmit Idle work together with Link Limbo feature. With the Transmit Idle feature enabled, if the link fails, the SRDF pair state will change to 'TransIdle' after the link limbo time expires (it depends on the time value you set 0 - 120 secs) instead of partitioned state.   Thanks

VMAX

SRDF/A "Partitioned" vs "TransIdle"

Was this post helpful?