I have and FCIP link to a remote site.
A 9509 on the HQ side and a 9216 on the DR side. I have a DS3 pipe and the fcip profile is configured for:
fcip profile 1
ip address 10.1.1.1
tcp max-bandwidth-mbps 1000 min-available-bandwidth-mbps 30
tcp send-buffer-size 5000
I am seeing tcp retransmit failures when a ton of data tries to go through the link.
I want to modify this profile on both ends of the connection and just had these questions:
1. Since I am seeing only these lines in the config, is everything else I see in the "sh fcip profile 1" command default and I do not have to be concenred with changeing anything else?
My concern is something else has been tweaked and I do not see it in the config, and the fcip link will not come back up after making these changes.
Will the link stay down until both ends have been modified?
I had a similar problem where our FCIP profile got changed at one end. Until Cisco Support talked me through re-creating the profile from scratch did the link come back up & working. All the Support had was "Tech-Support", so he must have got the info from that.
I don't know exactly what else was tweaked but in general, your maximum bandwidth command must be changed if it is not truly 1000 or using WAAS. A best practice would be shut the FCIP tunnel down on both sides and make the change and bring it up. The max bandwidth should be 45MB/sec and not 1000. To get the correct RTT, us the ips measure-rtt x.x.x.x interface
yeah you def. need to change the max bandwidth. just b/c your GigE interface can push 1000 doesn't mean your link can push that much.
the fun part is trying to figure out how much to give for FCIP without starving the other ethernet traffic if this is a shared link.
Thanks for the reply.
I am not sure I understand your coment on the shared link.
Even if this is a shared link, wont this change only affect the fcip traffic and nothing else?
Or are you talking about what I change the fcip interface to will have to take into account the other traffic, so I need to take the DS3 (45Mbps)and subtract whatever traffic there is and the rest is what I should set the fcip interface?
for instance if you have a 45Mbps link and you set your tcp-max to 45 Mbps and for an hour straight it needs to replicate a ton of changed tracks or data it is going to try and use the whole DS3 for replication b/c it is set to use the whole DS3(45MBPS). however if there is other ethernet traffic, say for instance SQL log shipping at the same time, it also needs to ride the DS3. so now your FCIP profile thinks it can go up to 45Mbps, which it will try if it is available, when in reality you need to save some for other ethernet traffic as well. when traffic starts to contend with each other you will tend to see write-latencies if any devices are setup in synch. replication or replication will suspend if it is set in async.
we have a link that is 155MB in total. we have 2 separate independant fabrics for redundancy. on each FCIP profile we set the tcp-max trans to something like 60MB. so 60MB+60MB=120MB for replication. that leaves 35MB for other ethernet traffic like (MSCS talk, heartbeat talk, etc...). by setting the limit to 60MB on each fabric or 120MB as a whole for replication, there is at least 35MB guaranteed for the rest.
there is a way to measure out what your tcp-max should be but not sure what the command is off the top of my head. i think 'tblancha' had it stated up a couple of replies. the .pdf's on cisco's site tell you how to run through setting up a FCIP profile and setting the Max and Min transmit values.
Thanks for the reply.
I have not ever modified and fcip profile.
This link is very importand and I was wondering about changing the max bandwidth:
How risky is making this change to the profile and it coming right back up?
Is this a relatively minor modification on the scale of just changing an ethernet interface?
What does the MDS do with all the data going into the Gigabit ethernet interface once I tell the fcip profile it has a max of 40M?
Is it buffered in the switch?
The MDS is trying to deliver data at the rate you configured that it could which is 1gig. But you don't have that much bandwidth. So, the current question is what is the MDS doing with the traffic when it's been told to do line rate but truly does not have that much bandwidth. And the result of that question is that you are getting drops and retransmissions. This change will bring down the FCIP tunnel so after you make the change, you will want to shut and then no shut the fcip interface. Should take maximum of 5 minutes. But it will cause an outage over the link.
Thanks for the reply.
Actually I was asking about the switch once I make the change to the fcip profile, the ethernet interface on the MDS still has a 1Gbps LAN capability.
What does the MDS do with the data transfered from my server into the switch Gigabit ethernet interface while the new configuration of the fcip link is now 40Mbps?
Does the switch buffer this data until the fcip can send it?
Do you have some experience using WAAS to optimize FCIP?. Actually we have a FCIP between two sites linked by 2x STM-1 (300 Mbps). We configured data compression and write acceleration, also we use jumbo frames, but we have an average rate of 2.2. Then, we would like to implement a couple of WAE to reduce the link to 1x STM-1. Do you think WAAS, could improve our rate better than 4 times?.