performance problem with mds9216i SAN

[g.sprietsma]

i have the following question. I want to exclude the san from the performance issue so i checked the interface errors en discards but no errors and discards found(or few). Further i checked in Device Manager the FC Enabled ports with the monitor option. I have checked the tabs Traffic, Protocol, Discards, Link Errors, Frame errors, Class 2 traffic, Class 2 errors. I have no clue what is normal under these errors and what definitely not.

does anyone have an idea ?

I have seen a lot of InvalidTXwords under Link errors. Is this normal ?

Hope anyone can help me ?

[inch]

G'day,

You really shouldnt see to many errors on switch ports. Try clearing your counters and then performing your tests again.

This will confirm that the errors are actually occuring during your tests.

See below for the output of a port without any errors...

You will see errors if you have bad cables and/or people have been yanking them in and out.

hope this helps!

sh int fc1/4

fc1/4 is up

Port description is ds-0-hb1-c4-p1

Hardware is Fibre Channel, SFP is short wave laser w/o OFC (SN)

Port WWN is 20:04:00:0b:5f:a3:c6:40

Admin port mode is FX

Port mode is F, FCID is 0x940001

Port vsan is 1000

Speed is 2 Gbps

Transmit B2B Credit is 101

Receive B2B Credit is 12

Receive data field Size is 2112

Beacon is turned off

Peer is type 002107 model 921 manufactured by IBM

5 minutes input rate 29859472 bits/sec, 3732434 bytes/sec, 2941 frames/sec

5 minutes output rate 64539872 bits/sec, 8067484 bytes/sec, 4560 frames/sec

107128878003 frames input, 196796010450408 bytes

0 discards, 0 errors

0 CRC, 0 unknown class

0 too long, 0 too short

41652147945 frames output, 69432378411836 bytes

0 discards, 0 errors

0 input OLS, 0 LRR, 0 NOS, 0 loop inits

0 output OLS, 0 LRR, 0 NOS, 0 loop inits

12 receive B2B credit remaining

101 transmit B2B credit remaining

[g.sprietsma]

Thank you for your reply

I also don't see any errors and discards on the interfaces.

But i see some errors under Device Manager and then Interface -> Monitor -> FC Enabled ..

Under Link errors i see a lot of TXInvalidWords

Under Protocol i see a lot of TxWaitCount and BBCreditTransistionFromZero which with my knowledge does indicate that the

What is your experience with these errors ?

[tblancha]

InvalidTXwords mean that we are receiving corrupt frames into the switch and the transitionfrom0 means that the switch had gotten down to having 0 BB credits left to send data to that device since it was not replenishing the BB credits. More than likely that is because there is a physical layer problem here and the data is getting mangled. Start with replacing the cable, then the sfp's, and then the HBA. Do one at a time and give it some time and monitor these 2 statistics to see if they decline.

[g.sprietsma]

thanxs for your reply. i will check if they are of the past or that they grow during time (which means the problem is still going on)

can i reset the counters ? I trieded it with clear counters interface all which didn't reset this counters.

i can use cumulative but this is realtime which isn't really handy.

[g.sprietsma]

is receiving corrupt frames the only way the switch can be out of bbcredits. is it wise to adjust the bbcredits ?

another question:

are there other counters we should watch to see if anything goes wrong. because no errors and discards on the interface does'nt mean a lot.

[inch]

Hi,

You can run out of buffer credit when frames are "lost" or when you run FC over long distances.

Cheers

Andrew

[g.sprietsma]

what do you mean with frames are lost. in my opinion out of bbcredit means that the buffer is full because the frames can be switched for some reason (long distance). where do you mean the frames are lost ?

[inch]

Howdy,

Frames can be lost if you have a faulty n/f_port.

The initial number of buffer credits is assigned when two N/F_ports login to each other. Each port will tell the other how much buffer credit it has available.

For example, port A logs into port B. Port A can handle 8 frames and port B can handle 100 frames.

Port A will set its BB_credit to 100 and port B will set its BB_credit to 8.

"out of bbcredit" means that the transmitting N/F_port will not send any more frames until it reveives a R_RDY from the receiving N/F_port.

So, if the n/f_port is over a long distance or broken and refuses to issue a R_RDY, you will run out of buffer credit.

I hope this helps!

Cheers

Andrew