I have absolutely no idea why, but this has just started working... will dig a bit deeper and try to find out
when you installed pxe, did you specify the option that dhcp was on the same box?
your dhcp server has the PXEClient option 60 set in it?
and in /usr/nsh/br/blpxe you do not see:
I've just discovered something odd...
PXE and DHCP are currently running on the same server. However, when I remove -Dblx.bmlistendhcp=true from blpxe I get the error PXE-E78 Coult not locate boot server.
It's only when I add it back in that it works, but then I cant start PXE while DHCP is running because I get "Address already in use" error in the pxesvr.log, so I have to start PXE first then DHCP and it works.
after you remove that entry, did you restart dhcp and pxe? and you have the PXEClient option 60 set in dhcpd.conf ?
that should allow pxe and dhcp to 'share' the port they both want
I just removed the -Dblx.bmlistendhcp=true entry from /usr/nsh/br/blpxe and stopped both DHCP and blpxe.
The entry option vendor-class-identifier "PXEClient"; is present in /etc/dhcpd.conf.
I stopped both services and then restarted them. Although the BM client now seems to attempt to PXEboot, it fails with the message...
PXE-E78: Coult not locate boot server
I just found that if I add the -Dblx.bmlistendhcp=true bit back into /usr/nsh/br/blpxe stop both DHCP and blpxe and then start blpxe first followed by DHCP my BM client now boots and find the boot server no problems.
This is making my head hurt :(
If I remove the -Dblx.bmlistendhcp=true option from /usr/nsh/br/blpxe should I be specifying an additional option within DHCP, something like next-server?
Is option 60 the vendor-class-identifier one, or is it something else?
option vendor-class-identifier "PXEClient";
-Dblx.bmlistendhcp=true is supposed to be set when the pxe is standalone (no dhcp)
try starting in the reverse order, dhcp, then blpxe...
is there anything in the pxesvr.log when the target fails to boot ?
If I start DHCP first then PXE, PXE states in the log that the "Address is already in use" and stops again. There's nothing in the log interesting when the client fails to boot with the PXE-E78 Coult not locate boot server error. I have logging level set to INFO, because with DEBUG there was just too much information.
So it seems that this is working with an incorrect option set in /usr/nsh/br/blpxe right? My DHCP and PXE are on the same server, so I should remove the -Dblx.bmlistendhcp=true from blpxe....but when I do it I get the PXe-E78 error...
I'm a bit worried now, because the DHCP running on the PXE server is only a temporary measure until our network guys can get their Lucent QIP system running DHCP within this VLAN... and then I'll have to change it to be setup for DHCP and PXE on different physical machines.
well, it should work fine if it's standalone, w/ that option set, no dhcp, so w/ the qip it should be ok.
that option should only be there if the box is standalone. on my VM (w/ prov working) i don't have the -Dblx.. set.
try turning on debug, boot the server and then stop pxe after it fails and post the log.
OK, thanks Bill... I will see where I'm at tomorrow... it's pretty late here, so will leave this until the morning. I think the QIP setup should be done first thing in the morning, so I'll probably be trying to setup PXE to work with that.
We shall see.
We made the changes to DHCP today on our provisioning VLAN, with our networks team adding a physical connection from their QIP server into this isolated LAN. They setup a DHCP scope and we began to test.
We were not able to pxe boot ANY bare metal clients. After a lot of diagnosis, it seems that the problem is that their DHCP server is not broadcasting it's replies, so while it leases an IP to the BM client, the PXE server doesn't hear the response from the QIP server and thus doesn't send the client the extra DHCP options it needs. This resulted in us getting continuous "No boot filename received" messages, or something like that anyway.
So, for now we've switched back to a Linux DHCP server running on the same physical box as our PXE server.
I can PXE boot clients now, no problem and can provision systems... that is all except one...
We currently have four test boxes...
HP ProLiant BL465c Blade - WORKING
HP ProLiant DL380 G5 - WORKING
HP ProLiant DL385 G5 - WORKING
HP ProLiant DL380 G3 - NOT WORKING!!!
For some reason, the DL380 G3, which has worked before is failing to PXE boot every time, with the message "PXE-E55: ProxyDHCP did not respond on port 4011".
I really don't understand this as all the other machines works OK, but this one has worked in the past.
I've updated the firmware on this box to the latest version to no avail. I've deleted it from PM, delete via BLCLI and even delete the database row for this device from PM_DEVICE table but nothing has helped... it still keeps on coming up with the same error. I don't understand it at all.
Message was edited by:
Message was edited by:
the dl380g3, it's on the right network and you have pxe enabled on the right nic ?
that's strange about the QIP, i thought we'd had that working ?
Are you still having this error with the PXE server? If so, I found this
article that might help you out: http://support.microsoft.com/kb/259670 .
Turns out that if you have the PXEClient option defied, it causes problems
with the DHCP server on Windows. If you remove that option, it might help
Hope this helps,
Sorry, I should have mentioned, that our servers are running Linux not Windows.
Anyway, this is now resolved after raising a support ticket and working on the problem along with BladeLogic support for about a week...
Turns out, that it was simply the settings is_use_broadcast = 1 that was causing all of our problems.