Ask Your Question

Why the metadata reuqest goes via a long way rather than given by Nova directly?

asked 2020-05-27 11:02:27 -0500

qlzhang gravatar image

Dear folks,

I have a question about the metadata, maybe it's very straightforward for you to answer but it's stuck in my head.

While checking the flow of how an instance accesses the metadata by using, it's quite a long path for the http packet to finally reach to the metadata agent then hand off to Nova, then I am wondering, since Nova is already the one who manages the instance, it knows well about the Instance ID, network ID, then why not we just let the Nova give the metadata directly to the instance?

Thank you and best regards qlzhang

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted

answered 2020-05-29 03:16:50 -0500

qlzhang gravatar image

Thank you Bernd for your help, agree with you that AWS uses this and it's a de facto and every cloud image follows.

I am just trying to understand, as the metadata is in the DB, then nova could easily read it and give to an instance locally rather than a long way going. By reading and thinking further, I suppose that the local way maybe not be agile and flexible enough as it can only be added at the booting-up time(am I right here?).

With the RESTful API call method, likely the instance can access it anytime, though I am not sure there is any real scenario that the instance gets the metadata at day2.

Best regards, qlzhang

edit flag offensive delete link more


Yes, and Nova does give it to the instance locally via the config drive.

I have an instance without ssh access that I communicate with by changing values of metadata items. No need for other communication channels. This is for training purposes, but could be used for production services as well.

Bernd Bausch gravatar imageBernd Bausch ( 2020-05-29 04:32:40 -0500 )edit

But in any case, why should Nova disable a standard means of transferring metadata? I am sure there are many applications out there that rely on the metadata API. Why break them?

Bernd Bausch gravatar imageBernd Bausch ( 2020-05-29 04:33:49 -0500 )edit

thanks Bernd, actually I am just checking and learning the flow of an instance booting up, then comes to this metadata part, dont' mean to break them, instead, curious that how it works and the motivation/necessity behind it.

qlzhang gravatar imageqlzhang ( 2020-05-29 11:44:07 -0500 )edit

answered 2020-05-27 20:18:26 -0500

Nova can make metadata available via the config drive, but the metadata API is an Amazon Web Services standard and therefore an industry standard. Also, I am not sure if cloud init supports the config drive.

edit flag offensive delete link more



We use cloud-init in combination with config-drive on a regular basis (Ocata, to be upgraded to Train) on openSUSE servers.

eblock gravatar imageeblock ( 2020-05-28 04:35:07 -0500 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2020-05-27 11:02:27 -0500

Seen: 32 times

Last updated: May 27 '20