Brocade Fibre Channel templates.

[pirx]

Can anynone tell me what I've to do to get the proper graphs with the Bytes/Second template.

During a benchmark the Brocade 200E Switch shows me this numbers (for the time of the benchmark, ~5 min dd read from raw disk)

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total
======================================================================================
289m 290m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 580m

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total
======================================================================================
294m 296m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 591m


But the graph shows only ~80-90M.

What do I have to chose when cacti ask me this:

Data Source [Summary of RX & TX Word]
A CDEF (math) function to apply to this item on the graph.

Currently I choose none.

[ame]

choose the "translate brocade words into bytes"-CDEF
It sums the RX and TX words and multiplies it with 4 - so you get the throughput bytes.

Change this in the graph template for the first entries of "throughput bytes" and cacti does not ask anymore....

But when I look to your graphs it seems correct....

Alex

[fosforo]

Hi all,

I'm trying to monitor trafic on brocade switchs using cacti but I get wrong values for trafic over 54 MBytes/second.
I think it's normal because we are using 32 bits SNMP counters so with a 300 seconds step we should have :

((2^37)-1)/((2^23)*300) = 54,61 MBytes/second

Is this correct ?
Does anyone solved this issue ?

((2^37)-1)/((2^23)*300) = 54,61 MBytes/second
Could you please explain the terms of this formula?

Anyone solved this problem? Cactid could be scheduled to run each 3/4 minutes instead of 5?

Thank you all

Fabio

[pirx]

choose the "translate brocade words into bytes"-CDEF
It sums the RX and TX words and multiplies it with 4 - so you get the throughput bytes.

Change this in the graph template for the first entries of "throughput bytes" and cacti does not ask anymore....

I tried this before, but the graphs didn't reflect the number I saw with portperfmon on the Brocade switch.

But when I look to your graphs it seems correct....

Hm, maybe I'm expecting something wrong.

If I make a test with dd for about 10 minutes and portperfmon shows constantly ~270m during this time, how can 80,51 M (bytes) as the max value in the graph be correct?

[ame]

which values did you monitor on your brocade? bytes/sec? words/sec?.....

We have here many really fast IBM storage boxes but none of them reaches 290 MB/sec.....

So first you should clarify which values you are monitoring on your brocade

Alex

[pirx]

which values did you monitor on your brocade? bytes/sec? words/sec?.....

I'm using the modified scripts from page 3 of this thread (I don't know how to link directly to the post):

PostPosted: Wed Sep 21, 2005 9:15 am Post subject: Reply with quote
Updated Host Template for Brocade Switches.

See that attached screenshots.

In the graph screenshot you see the throughput for a copy job that transfers data between port 5 and 6 (server and storage) with an average throughput of 70 MB/s. But the graph only show 16,98M as max. value.

We have here many really fast IBM storage boxes but none of them reaches 290 MB/sec.....

So first you should clarify which values you are monitoring on your brocade

The throughput of our RAID is only limited by the FC Controller. The RAID is capable of ~380 MB/s reading and ~170 MB/s writing (4 x 12 disk RAID 6 arrays in one 48 disk shelf).

[ame]

pirx,

I wanted to know what you have measured at your own benchmark test..

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total
======================================================================================
289m 290m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 580m

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total
======================================================================================
294m 296m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 591m

the values above - are they MB/s, Bytes/sec, words/sec....?

Try to make a test with IO-Meter (free downloadable)....after that you can compare the results much better....

Alex

[pirx]

pirx,

I wanted to know what you have measured at your own benchmark test..

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total
======================================================================================
289m 290m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 580m

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total
======================================================================================
294m 296m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 591m

the values above - are they MB/s, Bytes/sec, words/sec....?

This is output of the portperfshow cli command on the brocade switch.

From the manual:
portperfshow - Displays port throughput performance in bytes, kilobytes, or megabytes.

So the above should be 294 MB/s for a single port, which makes sense.

Try to make a test with IO-Meter (free downloadable)....after that you can compare the results much better....

I use tiobench for the test or a simple dd for reading from the raw device.

tiobench results:

Unit information
================
File size = megabytes
Blk Size = bytes
Rate = megabytes per second
CPU% = percentage of CPU used during the test
Latency = milliseconds
Lat% = percent of requests that took longer than X seconds
CPU Eff = Rate divided by CPU% - throughput per cpu load

Sequential Reads
File Blk Num Avg Maximum Lat% Lat% CPU
Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
2.6.18-5-amd64 20000 4096 1 387.17 49.92% 0.029 188.54 0.00000 0.00000 776

Random Reads
File Blk Num Avg Maximum Lat% Lat% CPU
Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
2.6.18-5-amd64 20000 4096 1 2.82 0.504% 4.156 48.27 0.00000 0.00000 558

Sequential Writes
File Blk Num Avg Maximum Lat% Lat% CPU
Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
2.6.18-5-amd64 20000 4096 1 180.30 40.65% 0.045 6851.57 0.00016 0.00000 443

Random Writes
File Blk Num Avg Maximum Lat% Lat% CPU
Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
2.6.18-5-amd64 20000 4096 1 1.16 0.496% 0.016 0.06 0.00000 0.00000 234

[ame]

pirx...I remember that we had a similar problem with running cacti (0.86) under debian. I thought that our SAN has nothing to do - peaks until 110 MB/s. Then I tried for updating reasons CactiEZ (0.87) installed on a VM and imported my Brocade templates. Switch to 1 minute polling and lay back....

Alex

[gandalf]

pirx...I remember that we had a similar problem with running cacti (0.86) under debian. I thought that our SAN has nothing to do - peaks until 110 MB/s. Then I tried for updating reasons CactiEZ (0.87) installed on a VM and imported my Brocade templates. Switch to 1 minute polling and lay back....

Alex

This reminds me of the usual 32bps boundary (COUNTER32 vs COUNTER64) when running 5 min polling ...
Reinhard

[trevorg]

Brocade uses 32bit counters. So yes you will hit the limit if you have heavy traffic in a 5mins period.

I got around this by using frames * 2112 (2112 is the size of the data portition of the fc frame). The only problem here is the FC spec says the data payload is variable but i have yet to see any issues in my testing.

Here is what i put in mrtg but im working on cacti side of things very soon.

1.3.6.1.4.1.1588.2.1.1.1.6.2.1.14.2&1.3.6.1.4.1.1588.2.1.1.1.6.2.1.13.2:public@x.x.x.x * 2112

[gigglesworth]

Thanks for these templates.

I notice your templates include a Graph Template called "Interface - Traffic - In/Out and Errors/Discards". Does this graph template intentionally work with regular Ethernet Interfaces?

I accidentally applied this Graph Template to some regular Ethernet Interfaces, and I get actual graphs. However, I'm not sure if the values in these graphs are valid.

I was wondering what the heck the labels "From CPE" and "To CPE" were doing there .

[lozanol]

We have 4 GB Brocade Switches and i have also the problem with the 32 bit counters.

I'm a cacti newbie, and i don't know how to change from tx, rxwords x 4 --> rx, txframes x 2112

I have try it, but every time i must reload the template. Can someone help me ?

Luis

[tomer]

hi all,

it seems that i have exactely the same problem like lozanol. i would like to use cacti to monitor the backup san using cacti. the san contains only 4-gbit san switches.

basically it seems that cati is working with the scripts i found here but i'm pretty sure that the values which are reportet by cacty are false. looking at the graphs i can see that we have a maximum of 80Mbit traffic on the switch ports. if i'm looing at the switches directly i can see much higher valuers. also looking at the backup application i can see that there is much more traffic on the switchports (if some DB backups are running we have up to 350MB/s of data on some swicth ports). i know that cacti onyl displays some average values but the backups jobs are running for hours and i still won't get any values higher than 100Mbit reported.

does anyone have similar problems with monitoruing 4gbit san ports ? is there a known way how a 4gbit san port can be managed ?

thanks for any feedback,
tomer

[trevorg]

If you dont know how to get around the issue the "simple" way would be just use the existing brocade template. Rip everything out of the FC switch graph template except for RX/TXframes and then place a CDEF on the graph items to perform FRAMES * 2112, which will give you bytes. So far this has worked very well for me.