[one-users] econe-run-instances idempotent?

Parag Mhashilkar parag at fnal.gov
Wed Oct 15 09:34:18 PDT 2014


That won't help. If the middle ware like HTcondor or run-instance commands starts putting anything in the user data, users are deprived of the functionality of user data.

AWS achieves idempotency with --client-token option, which to me seems like a tagging the request from the client side.

http://docs.aws.amazon.com/AWSEC2/latest/CommandLineReference/ApiReference-cmd-RunInstances.html

 
Thanks & Regards
+==========================================================
| Parag Mhashilkar
| Fermi National Accelerator Laboratory, MS 120
| Wilson & Kirk Road, Batavia, IL - 60510
|----------------------------------------------------------
| Phone: 1 (630) 840-6530 Fax: 1 (630) 840-3109
|----------------------------------------------------------
| Wilson Hall, 806E (Nov 8, 2012 - To date)
| Wilson Hall, 867E (Nov 17, 2010 - Nov 7, 2012)
| Wilson Hall, 863E (Apr 24, 2007 - Nov 16, 2010)
| Wilson Hall, 856E (Mar 21, 2005 - Apr 23, 2007)
+==========================================================

On Oct 15, 2014, at 11:27 AM, Daniel Molina wrote:

> 
> 
> On 15 October 2014 18:18, Parag Mhashilkar <parag at fnal.gov> wrote:
> Hi Daniel,
> 
> Let me rephrase what we meant.
> 
> As Steve mentioned in one of his emails, we use HTCondor to launch VMs. There is always a possibility that something can go wrong after a run-instance command is issued and before the ec2 server gets back to you with a valid instance id or an error message. Note that in this case the server may serve the request, but there is no way for the client to know the exact instance name that resulted from the request. Life is much easier if the client crashes after it gets back the instance name.
> 
> One way for the client to know the exact status of the request even after a crash (and before it gets the instance id) is to assign a unique identifier of its own like a tag name or something. This way client will never lose track of the request and result in the leaked VMs.
> 
> The way HTCondor is implementing this is by creating a new key pair and giving it a unique name and tracking the requests based on the key pair. But then we are hit by other limit ~300 that we observed.
> 
> Ideally, we would like to use the --keypair option and use the existing pre registered key. But than we cannot track the results of the create-instance in case something goes wrong before the instance id is issued.
> 
> And what about including a tag in the UserData instead of using the unique keypar?
> 
>  
> 
> Thanks & Regards
> +==========================================================
> | Parag Mhashilkar
> | Fermi National Accelerator Laboratory, MS 120
> | Wilson & Kirk Road, Batavia, IL - 60510
> |----------------------------------------------------------
> | Phone: 1 (630) 840-6530 Fax: 1 (630) 840-3109
> |----------------------------------------------------------
> | Wilson Hall, 806E (Nov 8, 2012 - To date)
> | Wilson Hall, 867E (Nov 17, 2010 - Nov 7, 2012)
> | Wilson Hall, 863E (Apr 24, 2007 - Nov 16, 2010)
> | Wilson Hall, 856E (Mar 21, 2005 - Apr 23, 2007)
> +==========================================================
> 
> On Oct 15, 2014, at 11:02 AM, Daniel Molina wrote:
> 
> >
> >
> > On 15 October 2014 17:57, Steven Timm <timm at fnal.gov> wrote:
> > If you call the same CreateInstances command more than once
> > is there any way that it will create the instance twice or not.
> >
> > If you call the command twice it will create 2 instances. You can also provide de MaxCount param in the CreateInstance command to create more than one instance at once.
> >
> >
> > Steve Timm
> >
> >
> >
> >
> > On Wed, 15 Oct 2014, Daniel Molina wrote:
> >
> > Hi,
> > What do you mean with "idempotent"? As long as the client implements the ec2 API, it should work
> > Cheers
> >
> > On 13 October 2014 20:33, Parag Mhashilkar <parag at fnal.gov> wrote:
> >       Hi,
> >
> >       We are using HTCondor to launch VMs in OpenNebula using ec2 interface and would like to know if the
> >       submit calls are idempotent.
> >
> >       Thanks & Regards
> >       +==========================================================
> >       | Parag Mhashilkar
> >       | Fermi National Accelerator Laboratory, MS 120
> >       | Wilson & Kirk Road, Batavia, IL - 60510
> >       |----------------------------------------------------------
> >       | Phone: 1 (630) 840-6530 Fax: 1 (630) 840-3109
> >       |----------------------------------------------------------
> >       | Wilson Hall, 806E (Nov 8, 2012 - To date)
> >       | Wilson Hall, 867E (Nov 17, 2010 - Nov 7, 2012)
> >       | Wilson Hall, 863E (Apr 24, 2007 - Nov 16, 2010)
> >       | Wilson Hall, 856E (Mar 21, 2005 - Apr 23, 2007)
> >       +==========================================================
> >
> >
> >       _______________________________________________
> >       Users mailing list
> >       Users at lists.opennebula.org
> >       http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
> >
> >
> >
> >
> > --
> > --
> > Daniel Molina
> > Project Engineer
> > OpenNebula - Flexible Enterprise Cloud Made Simple
> > www.OpenNebula.org | dmolina at opennebula.org | @OpenNebula
> >
> >
> >
> > ------------------------------------------------------------------
> > Steven C. Timm, Ph.D  (630) 840-8525
> > timm at fnal.gov  http://home.fnal.gov/~timm/
> > Fermilab Scientific Computing Division, Scientific Computing Services Quad.
> > Grid and Cloud Services Dept., Associate Dept. Head for Cloud Computing
> >
> >
> >
> > --
> > --
> > Daniel Molina
> > Project Engineer
> > OpenNebula - Flexible Enterprise Cloud Made Simple
> > www.OpenNebula.org | dmolina at opennebula.org | @OpenNebula
> 
> 
> 
> 
> -- 
> --
> Daniel Molina
> Project Engineer
> OpenNebula - Flexible Enterprise Cloud Made Simple
> www.OpenNebula.org | dmolina at opennebula.org | @OpenNebula

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2204 bytes
Desc: not available
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20141015/fc42b6d1/attachment-0001.bin>


More information about the Users mailing list