PXE boot issue

[Oaktech]

Morning chaps...

 

I have an issue that came to light late last week,

 

we have a new site that is connected to our main site via a 400m OM4 fibre, the connection goes from one HP 5412 to another, they are both the same revision of firmware.

 

We have some weird issues.

 

Weird issue 1: The PC's perform some operations very slowly, even machines which have previously been of a normal speed on our main site, seem very slow to start and log on when they move over there, in particular when XP reaches the 'applying security settings' and 'running startup scripts'

 

Weird issue 2: Our Unifi Pro APs won't adopt without several hard resets. The units adopt fine on our main site, but when we take them over to the new site they appear in the console, we adopt them, then they get stuck provisioning and require some times 3 hard resets before they work.

 

Weird issue 3: Machines get IP's from DHCP fine, unless you release the IP, then it sits for several minutes and says 'invalid directory' if you reboot the machine it picks up an IP with a fresh lease.

 

Weird issue 4: Machines won't PXE boot. Machines work fine on our main site and get a dhcp adress for netboot instantly but when you take them over to our new site they sit for 2 or 3 minutes and then say 'no dhcp or dhcpproxy offers recieved'

 

I'm pretty convinced this is all a switch config issue, something to do with broadcast i expect, but i'm blowed if i can find what it actually is. all of this is taking place on the same vlan.

 

Thoughts?

[Michael]

What speed is your OM4 connection operating at? I'm guessing 1Gbps or possibly more?

 

Typically it's the SFP that can create problems sending/receiving data down your cable. Are you using genuine or third party modules?

 

A good way of performing tests is to copy an ISO file (for example) from one server to another via the fibre. It shouldn't be unrealistic to expect 60-70MB/s or more connecting at 1Gbps using SATA drives. You should expect more with SAS drives.

[cullingsh]

Have you tried changing the igmp and/or spanning tree settings? Be the first place I would start - make sure they're using the same version of spanning tree too.

[Oaktech]

What speed is your OM4 connection operating at? I'm guessing 1Gbps or possibly more?

 

Typically it's the SFP that can create problems sending/receiving data down your cable. Are you using genuine or third party modules?

 

A good way of performing tests is to copy an ISO file (for example) from one server to another via the fibre. It shouldn't be unrealistic to expect 60-70MB/s or more connecting at 1Gbps using SATA drives. You should expect more with SAS drives.

 

The SFP's are genuine HP 10gbps units. I'm expecting it to be working somewhere less than 10 as it's slightly over transmission distance for 10gbps.

 

It moves a 165MB ISO in 3 seconds.

[twin--turbo]

400M is pushing MultiMode,

400M at 10GB is realy Pushing MultiMode to its limit.

 

Drop to 1GB FD, retest

Drop to 100MB FD, retest.

 

Rob

Edited December 2, 2013 by twin--turbo

[Oaktech]

400M is pushing MultiMode,

400M at 10GB is realy Pushing MultiMode to its limit.

 

Drop to 1GB FD, retest

Drop to 100MB FD, retest.

 

Rob

 

Ok, just hauled up the test results, the tester says it's 331.8m at it's longest and everything has passed the Fluke test as compliant to 10GbaseSR so I hope it isn't the fibre thats doing this...

[Oaktech]

Have you tried changing the igmp and/or spanning tree settings? Be the first place I would start - make sure they're using the same version of spanning tree too.

 

Well, they are both the same version of the FW and they both have identical settings for IGMP, obviously the STP looks different for each switch, but it's enabled...

[Michael]

331m is well within limits then, so this can't be the problem. I suspect it'll be a switch configuration issue. You could try restoring them to defaults and re-configuring them from scratch (out of hours).

[Oaktech]

Does anyone know how to extract a verbose config listing out of a HP ProCurve 5412ZL so I can compare them?

 

Resetting them both back to defaults its a non starter as we just have too much in one of them (it's our core, default gateway and routes our primary site subnet to us) to start again!

[Michael]

Does anyone know how to extract a verbose config listing out of a HP ProCurve 5412ZL so I can compare them?

 

Resetting them both back to defaults its a non starter as we just have too much in one of them (it's our core, default gateway and routes our primary site subnet to us) to start again!

 

Well you could start with the switch on the other end, rather than your core. Do you have multiple fibre links like this, or just one the one?

[Oaktech]

Well you could start with the switch on the other end, rather than your core. Do you have multiple fibre links like this, or just one the one?

 

There is just the one link like this. Its a 24core fibre so I have more than one core to use, but I can only use one at a time as it's my only set of SFPs and interconnects.

[Michael]

There is just the one link like this. Its a 24core fibre so I have more than one core to use, but I can only use one at a time as it's my only set of SFPs and interconnects.

 

Yeah 10Gbps SFPs are rather expensive compared to 1Gbps. It may be worth using 1Gbps SFPs as no doubt this'll work fine over your 331m fibre link. It also lets you test whether it is a speed/tolerance issue. It's pretty difficult to consistently use all the bandwidth 1Gbps has to offer in most instances.