This post is more than 5 years old
40 Posts
0
9868
August 6th, 2010 06:00
incomplete POST
I'm creating an object using the object interface. Normally I issue the POST then get back a esponse with the Location header containing the object identifier so that I can access the new object later. Suppose that there is a network issue where the POST succeeds, the object is created but I never get the response with the Location. An object now exists and I have no way of accessing it becasue I don't know the object ID. I can't read or modify it and I can't delete it. I will be charged for this unusable object forever.
Is this correct or am I totally missing something?
0 events found
No Events found!
rbala1
222 Posts
0
August 6th, 2010 08:00
That's certainly a possibility. I would suggest using listable tags so that you can access the object after it's created.
You can get a list of all objects with a given listable tag and manage it appropriately.
Have you see AtmosFox, the Firefox object browser? http://www.suchisoft.com/atmosfox/AtmosFox.xpi
madler1
40 Posts
0
August 6th, 2010 08:00
With the millions of objects that I generate I'm trying to stay away from using listable tags much. However, I supose I could generate a unique tag for every POST and search for that tag in the event that the POST fails.
Thanks
rbala1
222 Posts
0
August 9th, 2010 06:00
You're right --- we probably wouldn't want a unique tag per object, but an otherwise identifiable tag may suffice.
And it does indeed have to be listable to generate a list of objects based on tag.
rbala1
222 Posts
0
August 9th, 2010 06:00
Since Atmos Online is a reference architecture and development platform there is no billing enabled. It's an open platform for our developer community to use in the process of building their applications to support our partner cloud service providers and the on-premise deployments of Atmos.
But in reference to the concept of orphaned objects, Atmos doesn't necessarily know whether they are orphaned or not. The expectation with the object interface is that the application creating and managing objects needs to persist the Atmos object ID in it's own database in order to access to the object later. So we are unaware as to the state of storing and managing that identifier.
madler1
40 Posts
0
August 9th, 2010 07:00
I was going make the same point that Atmos can't be aware that an object is orphaned. Also, my aspect of the question was not necessarily related to Atmos Online but to the Atmos interface that other providers such as Synaptic are changing for.
But I am trouble by the suggestion that having millions of unique tags will lead to performance issues. I am aware that appliing the same listable tag to many objects (hundreds of thousands) will lead to performance issues when searching the tags. Is there any way that I can use listable tags with a large number of objects and not see a significant impact on performance. This includes delete performance which we make heavy use of.
I suppose that I can apply the listable tag when POSTing the file and then remove the tag from the object after I have recorded the object ID. But perfroming an additional operation (tag removal) is something I would much rather avoid. I would rather use that http request to transfer data than to perform house keeping.