Menu

#443 Some NID started services cannot be restarted by AMF

4.4.RC2
fixed
None
defect
osaf
-
major
2014-02-18
2013-06-05
No

AndersW had floated a patch for this - http://list.opensaf.org/pipermail/devel/2013-January/028573.html .
However this topic needs a little more consideration. This ticket/patch by itself isnot having any problems, however it opens up a lot of other things that were not possible without this patch like performing admin op on middleware SU.

One of them is, for AMF to allow the below commands only if they are initiated by SMF, i.e. during upgrade. For instance AMF could achieve this by seeing if a campaign is in progress(sumaintenance campaign attribute?)

amf-adm lock safSu=SC-2\,safSg=2N\,safApp=OpenSAF
amf-adm lock-in safSu=SC-2\,safSg=2N\,safApp=OpenSAF
amf-adm unlock-in safSu=SC-2\,safSg=2N\,safApp=OpenSAF

Also, additional considerations discussed in the below links have to be accounted for:
http://list.opensaf.org/pipermail/devel/2011-September/017903.html
https://sourceforge.net/p/opensaf/tickets/436/

Migrating this ticket for reference:
With the exception of immnd, the services started by NID cannot be restarted by AMF. This is because they always assume that they were started by NID, and therefore do not register with AMF.
The logic in immnd for handling start from either NID or AMF should be reused also in the other NID started services.

Changed 9 months ago by anwi

status changed from new to accepted
Changed 9 months ago by anders

Actually, the IMMD is also monitored by the AMF, but that monitorign can only start
after the AMF is up, which has to be after IMMNDs have loaded.
The problem should only exist for early starting services.
Changed 9 months ago by anders

And "early starting services" I suppose equals NID started services.
Changed 9 months ago by nagendra

Except IMMND, other services are not restartable services. Once any one of those services are down, node reboots.
Changed 9 months ago by hafe

milestone changed from 4.2.2 to future_releases
I removed the milestone for this ticket since we haven't concluded what this is about yet. As Nags say, an active NID started service cannot just be restarted if it crash. But there are other problems. NID does not supervise its started services more than one by one. It forget each one after it has responded. This is not good since when AMF starts any NID started service could be down and it is not really designed for that.
Another use case is that unlock-instantiation of the 2N opensaf SU does not work because of this. This use case could be useful to change the actual AMF model of opensaf itself.
Changed 7 months ago by anwi

priority changed from major to minor
type changed from defect to enhancement
patch_waiting changed from no to yes

Related

Tickets: #443

Discussion

  • Mathi Naickan

    Mathi Naickan - 2013-06-05

    Migrated from #2802

     
  • Mathi Naickan

    Mathi Naickan - 2013-11-10
    • status: unassigned --> accepted
     
  • Anders Widell

    Anders Widell - 2014-01-09
    • Milestone: 4.4.FC --> 4.4.RC1
     
  • Mathi Naickan

    Mathi Naickan - 2014-01-28

    Doc updates are pending. Nagendra to update the AMF PR doc.

     
  • Mathi Naickan

    Mathi Naickan - 2014-01-28
     
  • Mathi Naickan

    Mathi Naickan - 2014-02-07

    Moving this to 4.4 RC2 in preparation for RC1.

     
  • Mathi Naickan

    Mathi Naickan - 2014-02-07
    • Milestone: 4.4.RC1 --> 4.4.RC2
     
  • Mathi Naickan

    Mathi Naickan - 2014-02-18
    • status: accepted --> fixed
     
  • Mathi Naickan

    Mathi Naickan - 2014-02-18

    The patches are all pushed. High level README update is pending.
    Will do it via https://sourceforge.net/p/opensaf/tickets/792/.

     

Log in to post a comment.