cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
717
Views
0
Helpful
1
Replies

MTU

fadisodah
Level 1
Level 1

I have implemented 2x sites(SiteA+SiteB) SAN DS8100 storage
asynchronous replication over IP/MPLS Layer2 with 35Mbps speed(
SAN_A--FCIP_A---Router_A--IP/MPLS--Router_B--FCIP_B--SAN_B), When I
check the connectivity I see timeouts, also I have adjusted the MTU
from 1500 to 1392 after some measuring ping hostname -l 1392 -f
after reading http://www.cisco.com/en/US/docs/solutions/Enterprise/Data_Center/HA_Clusters/HA_FCI_4.pdf

-------/snapshot of the ping during SAN replication between 2
sites/-------------------
Reply from 172.20.0.105: bytes=1390 time=93ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=80ms TTL=124
Request timed out.
Request timed out.
Reply from 172.20.0.105: bytes=1390 time=76ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=93ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=85ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=82ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=93ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=94ms TTL=124
Request timed out.
Request timed out.
Request timed out.
Reply from 172.20.0.105: bytes=1390 time=93ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=93ms TTL=124
Request timed out.
Reply from 172.20.0.105: bytes=1390 time=93ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=78ms TTL=124
Reply from 172.20.0.105: bytes=1390 time=80ms TTL=124
------------------/End/------------------------
Please not that when I stop the replica the ping get stable without
any interruption...
I can post configs if you need it, the MTU adjustment did not fix the
problem of instable MPLS connectivity
Any help would be appreciated to adjust any paprameter of the FCIP
router, MPLS Router, Edge Router or any tunning parameter...etc.

Thanks

1 Reply 1

Marcin Latosiewicz
Cisco Employee
Cisco Employee

Hi,

I'm not sure if this is the correct section of forums I don't see how the problem could be related to MPLS itself.

It looks like during replication some buffers are exchausted. You are right that avoiding fragmentation could be one of the possibilities to aleviate.. if it's THE problem at all.

AFAIR FCIP will use TCP for transport and not raw IP? If so you can adjust MSS (ip tcp adjust-mss) on an L3 interface of Cisco routers.

MSS is typically MTU - 40.

Marcin