alloc failed

[earthgecko]

Hi

Just a quick one (two). I am sure there are fundament hardware virtualisation reasons for these, especially the first, however the second does not make sense.

Exceeding the resource limits so as to make HA on the Grid servers unavailable, even if this is just a few Mbps over the half the resources limit results in:

1) HA becoming unavailable - would it not be possible to reconfigure the application/s on the fly and reduce resource allocation proportionally to overcome this?

2) Not being able to restart a running component in an application on the Grid? Now this one is a little more frustrating, the first one I can understand as the resources are not available, however at the component level restart surely the application already has those resources assigned and the component was running with them with no HA available. Why does a component restart in this edge case scenario return:
Failed to start component APPLICATION:main.APPLIANCE - alloc failed - see log for details.

Why it does not make sense, is that without changing any resource use apart from restarting the application, resolves the issue, even though that very action IS going over the HA limit. Why is the application different to the component?

Thanks

[PeterNic]

Earthgecko,


1) HA becoming unavailable - would it not be possible to reconfigure the application/s on the fly and reduce resource allocation proportionally to overcome this?

While CPU and bandwidth theoretically could be changed dynamically, memory and disk may not be as flexible and may make the shrunk appliance/applications dysfunctional / performing below required levels. It is not, however, a bad approach and we will review it for future versions -- I am guessing that a "prioritization" or "rebalancing" appliance may monitor what's happening on the grid and make the adjustments according to some policy (which would be highly customizable). Another alternative in most cases is to be able to request an additional server from the hardware resource pool -- also something that will be in the works.

Note that AppLogic is an N+1 system -- so to have HA you need 1 server free, no matter how many servers are in the VPDC; not half.


Failed to start component APPLICATION:main.APPLIANCE - alloc failed - see log for details.

This one is a bug introduced in 2.4 -- there is a hotfix available for it; the hotfix can be applied without disrupting the grid's operation. The hotfix is hf2794, designated as a mandatory hotfix -- please contact your service provider to have it installed and sorry for the inconvenience.

Best regards,
-- Peter

[earthgecko]

Thanks once again Peter, I will get in touch with them.