DR tunnel missing ?

[Max Pierre]

Hi,

Hi,

The diagram above shows a logical view of the DR connections between EVAs
during normal hours.

We have every day around 03:30 connection losts between evas with DR start
logging for a few minutes. Problem comes from Site A, Ctrl A, FP3.

Error message in the eva is :
A Data Replication Group has transitioned to the Logging state
because the alternate Storage System is not accessible : 0c1b5f0c

What seems strange to us is the missing Active DR connection between
Site A CTRL A and site B CTRL B where only DR Available links are established.

Any idea ?
Thanks in advance


San switches are Brocade 8Gb SAN Switch for HP BladeSystem c-Class, v 7.1.0b
Eva P6300 with XCS 11001100
Luns are balanced between controllers Path x-Failover/failback.
No errors on brocade ports (FP or ISL)

Two fabrics,
all FP3 on fabric 1, all FP4 on fabric 2 for DR only
FP1 and FP2 used for applications only

Implementation as documented in the P6000 Continuous access implementation guide,
figure 38 Dual-Fabric zoning components.

[Jason_L]

Hello,

If I understand your question, you are asking why you go into a logging state when "problem comes from Site A, Ctrl A, FP3" and you have other ports  (on same fabric AND on a different fabric) that are configured to carry replication traffic?

I also assume that the diagram you provided was taken from the host port data replication settings found in Command View?

Can you post a screenshot of your host port data replication settings from Command View?

[linarvz]

Hi Max Pierre,

A tunnel failing at the same time at 3:30 every night in an otherwise stable setup would lead me to think of an external event (such as a backup or remote object job) affecting DR tunnel.

Could it be that some other activities start at that very time and overload either SAN fabric path or array ports?

Linar

[Jason_L]

To expand upon this more while we wait for a screenshot of the replication ports...

I agree with the other poster that other activities utilizing the site link can have impacts on the CA replication - such as backup.  I have persoanlly seen other instances where the target array was WAYY overutilized during certain parts of the day and latency with the DR Group volumes became very large, to the breaking point and the EVA's would lose communication.

For all ports to lose tunnel, i would look for something in common assuming the host ports / zones / etc,  are configured correcly.