UNSOLVED

stefan_varga

updated

8 years ago

0

3017

March 19th, 2018 06:00

lacp and dynamic pools

Hi,

We observed that in case of one physical interface goes down within LACP aggregation, dynamic IP are failing over to another node. Static pool IP addresses are staying UP on the same LACP aggregation.

My question is, if this expected behavior for Isilon OneFS.

Cluster: Gen6 H500

OneFS 8.1.0.2

Thank you,

Stefan

  • Yan_Faubert

    117 Posts

    1639

    0

    Posted March 19th, 2018 08:00

    This doesn't sound right, imagine you lose an entire switch, will you end up with all dynamic IP addresses getting moved?  I'm checking internally on this...

  • 1639

    0

    Posted March 20th, 2018 01:00

    yes, disabling one of the two physical interfaces in LACP aggregation per node, dynamic IP are moved to other nodes. After disabling the interface on last node the IP stay on the last node. SSIP gets moved no nowhere (disappears) which is a real issue.

    Support says, that dynamic IP are moved in case of one iface is offlined in LACP - which sounds for me like a bug, not mentioning SSIP behavior.

    Stefan

  • Yan_Faubert

    117 Posts

    1639

    0

    Posted March 20th, 2018 06:00

    I've confirmed with some internal resources that this issue will get escalated to engineering soon since it shouldn't do that.  This is a fairly new issue; if you have a case opened with support you can point them to this thread and they can contact me via internal email.

  • Yan_Faubert

    117 Posts

    1642

    0

    Posted March 22nd, 2018 11:00

    Ack, will do.

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    1642

    0

    Posted March 22nd, 2018 11:00

    Yan,

    Can you provide an update when you get one ? We are in the early stages of planning 8.1 upgrade and migration to Gen6 nodes.

    Thank you

  • Yan_Faubert

    117 Posts

    1642

    0

    Posted March 28th, 2018 06:00

    Issue being investigated by engineering as part of bug# 222296.

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    1642

    0

    Posted March 28th, 2018 10:00

    Thank you Yan. I guess it's too early to ask for a resolution ETA.

  • Yan_Faubert

    117 Posts

    1642

    0

    Posted March 28th, 2018 12:00

    Correct, no info on that yet.

  • akkornel1

    4 Posts

    1642

    0

    Posted April 4th, 2018 14:00

    That is interesting timing: I think we hit exactly the same issue last week: We have two Dell N-series switches in a stack.  We have LACP in use, so each node has one 10G on switch A, and one 10G port on switch B, and LACP in use (which works because they are stacked).

    The reboot of one switch caused one 10G link to drop on all nodes (the other 10G link stayed up), and caused an "All links in LACP offline" event, which caused all dynamic pool IPs to move away.

    And we are on OneFS 8.1.0.2 as well…

    Our SR is 10593111

  • Yan_Faubert

    117 Posts

    320

    0

    Posted April 5th, 2018 06:00

    From your SR, it sure looks like the same issue.  FYI, the previously mentioned bug was the escalation bug#.  The actual product bug# for the 8.1.0.x code branch is 222431.  What you can do is request a patch for 8.1.0.2 for that bug via support.  The bug isn't fully resolved yet but is being worked on.