euca-describe-availability-zones output is a hint, not totally accurate - and should be documented

Bug #610987 reported by C de-Avillez on 2010-07-28
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Eucalyptus
Invalid
Undecided
Unassigned
euca2ools (Ubuntu)
Low
Dave Walker
Maverick
Low
Dave Walker
eucalyptus (Ubuntu)
Low
Unassigned
Maverick
Low
Unassigned

Bug Description

This may be related to bug 610259.

During my test runs on Eucalyptus V2, I start from an idling system, and then 'euca-run-instances -n <value>'. The <value> is calculated as a random between 1 and <# of possible instances for this type / 2).

The very first euca-run-instances -- on this test -- was set to '-n 12', for type=c1.medium. All 12 instances were accepted for start. An 'euca-describe-availability-zones verbose' *before* the euca-run-instances reported 40/40 as Free/Max for c1.medium.

AFTER instance run submission, 'euca-describe-availability-zones verbose' reports 33/40 for type=c12.medium. Of course,

(40 - 12) != 33.

test log showing the sequence of events will be attached.

eucalyptus-cc 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
eucalyptus-cloud 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
eucalyptus-common 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
eucalyptus-gl 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
eucalyptus-java-common 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
eucalyptus-sc 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
eucalyptus-walrus 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed
libeucalyptus-commons-ext-java 0.5.0-0ubuntu2 eucalyptus-commons-ext install ok installed
uec-component-listener 2.0~bzr1211-0ubuntu1 eucalyptus install ok installed

C de-Avillez (hggdh2) wrote :

whole log attached; the issue can be seem throughout the log, but it is clearly visible on the very first runs.

C de-Avillez (hggdh2) wrote :

milestoning to Alpha3, setting as High

Changed in eucalyptus (Ubuntu):
importance: Undecided → High
milestone: none → maverick-alpha-3
C de-Avillez (hggdh2) wrote :

I ran another test -- I thought it might have something to do with 'euca-run-instances -nxx'. So, starting with a just-booted cloud, I ran a sequence of 'euca-run-instances -n10 ...' intermixed with 'euca-describe-availability-zones verbose'. I then stopped for lunch, and then ran two more e-d-a-z.

C de-Avillez (hggdh2) wrote :

On the log above -- it is interesting to note that suddenly we lost a whole NC (m1.small MAX went from 80 to 64) for a while. After lunch, the NC was back in, as misteriously as it had disappeared.

C de-Avillez (hggdh2) wrote :

full logs now available:

bzr+ssh://bazaar.launchpad.net/~hggdh2/%2Bjunk/uec-qa/
Pushed up to revision 24.

Thierry Carrez (ttx) on 2010-08-02
Changed in eucalyptus (Ubuntu Maverick):
assignee: nobody → Dave Walker (davewalker)
Thierry Carrez (ttx) on 2010-08-02
Changed in eucalyptus (Ubuntu Maverick):
milestone: maverick-alpha-3 → ubuntu-10.10-beta
importance: High → Medium
milestone: ubuntu-10.10-beta → none
status: New → Confirmed
milestone: none → ubuntu-10.10-beta
Changed in eucalyptus:
assignee: nobody → chris grzegorczyk (chris-grze)
Dave Walker (davewalker) wrote :

Upstream has indicated that "euca-describe-availability-zones verbose" is a hint of what the availability is, and not necessarily current. Providing this is not a regression from 1.6.2 (waiting on upstream to confirm), this requires a documentation fix. Therefore setting to low priority.

Changed in eucalyptus (Ubuntu Maverick):
importance: Medium → Low
status: Confirmed → In Progress
Dave Walker (davewalker) on 2010-08-13
summary: - euca-describe-availability-zones verbose incorrect output
+ euca-describe-availability-zones output is a hint, not totally accurate
+ - and should be documented
Thierry Carrez (ttx) on 2010-08-25
Changed in eucalyptus (Ubuntu Maverick):
milestone: ubuntu-10.10-beta → none
Thierry Carrez (ttx) on 2010-09-02
tags: added: server-mrs
Dave Walker (davewalker) wrote :

euca-describe-availability-zones is part of the euca2ools package, and is therefore invalid bug against eucalyptus. Opening task for euca2ools.

Changed in eucalyptus (Ubuntu Maverick):
status: In Progress → Invalid
assignee: Dave Walker (davewalker) → nobody
Changed in euca2ools (Ubuntu Maverick):
status: New → Confirmed
assignee: nobody → Dave Walker (davewalker)
Thierry Carrez (ttx) on 2010-09-14
Changed in euca2ools (Ubuntu Maverick):
importance: Undecided → Low
Dave Walker (davewalker) wrote :

I intended to get this in the next upload of euca2ools as it doesn't seem this is worthwhile of an upload on it's own, but this situation hasn't arisen. Marking won't fix for Maverick, and will open a task for Natty when available. Of course, in the unlikely event euca2ools needs an upload before Maverick release - this should be added.

Changed in euca2ools (Ubuntu Maverick):
status: Confirmed → Won't Fix
Thierry Carrez (ttx) on 2010-10-04
tags: removed: server-mrs
C de-Avillez (hggdh2) wrote :

There is something that has been bothering me for a while now: e-d-a-z verbose may be a "hint of what is the availability, but not necessarily current". That's OK, any measures on a dynamic system are by definition outdated at the moment of measurement.

The point is 'e-d-a-z verbose' should behave like -- say -- 'ps aux' does: the measurement is *correct at the instant it was taken*. If it *does* behave this way, then I do not see a need to document anything (which is to say, this bug should be set INVALID).

Now, the only thingie left is the sudden drop of maximum sessions (the equivalent of a full NC suddenly went AWOL). But (1) this would be another bug, and (2) I never saw this happening again.

My vote: close INVALID.

Dave Walker (davewalker) on 2012-02-15
Changed in euca2ools (Ubuntu):
status: Confirmed → Invalid
Changed in eucalyptus:
assignee: chris grzegorczyk (chris-grze) → nobody
status: New → Invalid
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers