6

We are currently using multiple HP ProCurve V1810-48G switches in our environment which we believe are causing issue with GPO based software installations. This problem surfaced after installing SSD drives into new laptops which seems to have sped up the boot processes enough to cause issues. We cannot reproduce this issue using HDD's.

It feels like the issue is related to Spanning Tree support. The computer appears to be making the connection long enough to satisfy the "wait for network" condition, but is being disconnected prior processing the GPOs and contacting the file share. joeqwerty commented that the root solution is to turn on PortFast or the equivalent, but this setting is non-configurable on the V1810.

I had found information on modifying "GpNetworkStartTimeoutPolicyValue" to specify a wait time before continuing to boot, however as soon as the network is detected the Timeout expires and booting continues. Modifying this value has no effect in testing. I tried inserting a startup script to ping localhost for a set amount of time to induce a delay as suggested in other places online, but that method also did not have the intended effect.

It seems the simplest solution is to upgrade the switches to a model that has this feature exposed but at this time that is not an option I have available.

Does anyone know of a workaround that could force the boot process to wait for the duration timeout even if the network is detected briefly?

Zac67
  • 13,684
Melikoth
  • 395
  • 2
  • 3
  • 15

2 Answers2

1

While this is an old question, the underlying problem is still present, with any kind of switches. There's no need to fiddle with the hosts' IP stacks though.

Instead of meddling with the symptoms, you should solve the root problem:

When you activate a spanning tree protocol on a managed switch, any switch port linking up goes through the listening state by default, where it doesn't forward any frames yet. This state is to guarantee that a new bridge/switch is incorporated into the spanning tree before actually allowing traffic.

On end-node ports that behavior often causes problems with DHCP, especially with host booting up fast. On those ports you should always activate the admin-edge-port option (portfast on Cisco gear), so the port skips the listening state and starts up in the forwarding state right away.

On the 1810v2 series, use Switching → Spanning Tree → Spanning Tree Port Settings → Admin-Edge → Edge. On Procurve switches, it's spanning-tree <port> admin-edge-port.

Auto Edge is not the same and doesn't solve your problem, as it only activates edge mode after BPDUs fail to be received.

Zac67
  • 13,684
0

Have you tried using Always wait for the network at computer startup and logon

It would be under: Computer Configuration/Policies/Administrative Templates\System\Logon

Please note that it will slow down the login process.

References: https://technet.microsoft.com/en-us/magazine/gg486839.aspx https://support.microsoft.com/en-us/kb/305293

Eugene R.
  • 11