Disqus - Latest Comments for cdent

Re: OpenStack Nova Scheduling Optimizations at CERN

cdent — Fri, 05 Jul 2019 11:59:18 -0000

It sounds like most of the opportunities for improvement are in nova-scheduler itself, either in the way it makes requests or the way it handles responses. Have you identified any areas on the placement service side that might be bottlenecks? For most of the rest of the Train cycle we'll be trying to focus a fair amount of energy on performance, so having some data on where to look would be great.

(Thanks, as ever, for writing this kind of thing up, I think it is _extremely_ useful.)

Re: Write – Mailing List Review

cdent — Mon, 26 Nov 2018 15:48:25 -0000

Apparently the new `openstack-discuss` list is not going to have a `Reply-To` header to the list because of DKIM signatures. If your email tool doesn't do Reply-To-List (`L` in mutt) then you could consider changing the message itself as it arrives. Here's a procmail rule:

```
# fixup reply to on openstack-discuss
:0fhw
* List-Id:.*openstack-discuss.lists.ope...
| formail -i "Reply-To: openstack-discuss@lists.openstack.org"
```

There are plenty of options. All more complicated than the mailing list server doing it, but there are reasons.

Re: Write – Some Opinions on OpenStack

cdent — Mon, 09 Jul 2018 13:26:41 -0000

I suppose it is a question of degree. While you don't have will etc to review and fix throughout all the bits, would you agree that you care about the success of OpenStack (whatever that is) as an overall thing? It certainly looks that way from the outside.

Also, to what extent do you think the social structuring of the projects and the architecture thereof enforces the way people identify with different thing in OpenStack? Would it be different if that structure was different?

I do, and in a roundabout way I guess my comment was a way of saying: If we need or want to see change in the way people identify with the overall project we need to consider changing how things are structured.

Where I think we (and I do mean you and me here, not the "we" the community, but probably them too) run into trouble is that because of constraints on available resources and the many forces that come together to cause goals, we are often called upon to dedicate ourselves to features rather than enabling others to create them. That's economics, I guess, and I wish we could make all the costs more clear.

Re: Write – OpenStack Casual Contribution

cdent — Wed, 22 Nov 2017 03:37:47 -0000

Doug Hellman pointed out a useful [video](https://youtu.be/VS6IpvTWwk... that distinguishes between first and second time contributors. A great deal of energy is spent on enable first time contributors but not all those people want to come back. If creating maintainers is a goal, then keeping the second time contributors is critical.

There's also a great deal of interesting stuff in their about sustaining open source.

Re: Write – Ocata Multi-Node Devstack Play

cdent — Thu, 22 Jun 2017 17:50:21 -0000

Sorry, I don't think this is something we're going to be able to figure out in the comments of a blog posting as there's going to be a lot of back and forth. Also I'm on holiday at the moment, so don't have much chance to look into things right now. I'd suggest you try contacting people in IRC or on https://ask.openstack.org/

Good luck

Re: Write – Ocata Multi-Node Devstack Play

cdent — Thu, 22 Jun 2017 04:48:43 -0000

Did you do `nova-manage cell_v2 discover_hosts` on the controller host?

Without more information it's really hard to say what's going on. Did the `stack.sh` on the compute node indicate that it had finished successfully?

Is there network connectivity between the 2 physical machines.

Re: Write – Remote Literacy

cdent — Thu, 17 Nov 2016 07:36:31 -0000

Thanks. Nice to know you're still out there reading.

Re: Write – Simple Resource Provision

cdent — Mon, 08 Aug 2016 11:12:46 -0000

I guess we'll just disagree, but now that we're off the mailing list about this I can dispute you without distracting the real work. ;)

In my mind there is a storage resource provider. It provides DISK_GB. It also provides SSD. That's all there is to it. Just like a compute node provides several different classes of inventory, so does the storage provider. I now have a constrained but sufficiently expressive grammar.

In your model the the grammar is less constrained resulting in something that is more difficult to reason about, more difficult to model in multiple solutions and requires a more complex representation in both the request and response.

Re: Is OpenStack Broken?… Not So Fast, My Friend

cdent — Thu, 07 Apr 2016 09:21:38 -0000

I really don't like the auto parts metaphor. It's useful for companies for OpenStack to be parts from which you can pick and choose to build a product but it's not useful individual users or small groups. So I wrote a response over here: https://anticdent.org/opens...

Re: Write – Unpacking the Nova Scheduler (part 3 of ?)

cdent — Mon, 28 Mar 2016 14:48:28 -0000

Ah yeah, thanks for catching that. Will update with a correction.

I'm glad you liked the series. Was very useful for me to write.

Re: Write – Persistent IRC Considered Harmful

cdent — Tue, 22 Dec 2015 06:27:32 -0000

Yeah. Usenet's attention to threading and tools for managing threads was great.

Re: Write – Persistent IRC Considered Harmful

cdent — Mon, 21 Dec 2015 10:49:55 -0000

Uh, read again. This article is not asking for full server logs. I'm saying that server logs are a bad place for important project information to be and that IRC bouncers help to enforce people leaving important information in IRC.

Re: Using Gabbi and Hypothesis to Test Django APIs | Wildfish

cdent — Thu, 15 Oct 2015 11:08:30 -0000

Eric, I suppose it could be useful for PMs and less technical people, but mostly it is for people who want to be writing the tests closer to HTTP rather than Python. The example dan gives in his response to you makes it pretty clear that the `self.client` being used is pretty clean. That's not always the case. If look at http://redhat.slides.com/ch... from a recent presentation you can see a more dramatic example of situations where gabbi is more direct than Python. That example is from OpenStack ceilometer.

dan, btw, in the next release (happening in the next few days) of gabbi `test_suite_from_yaml` is deprecated and the new more reasonably named method is `test_suite_from_dict`. This is part of a slow refactoring to make it easier to create tests from a variety of inputs.

Re: Using Gabbi and Hypothesis to Test Django APIs | Wildfish

cdent — Thu, 01 Oct 2015 07:09:14 -0000

Wow, this is awesome, thanks for writing this up. I've been hoping to explore this with gabbi since learning about Hypothesis.

Re: Installing nova-docker in N easy steps

cdent — Fri, 06 Feb 2015 06:35:36 -0000

Is this line right: "python --enablerepo=rawhide install python-six systemd", you want yum instead of python yeah?

Re: “Vanilla OpenStack” Doesn’t Exist and Never Will

cdent — Thu, 05 Feb 2015 06:48:16 -0000

It may be the case that there is no vanilla OpenStack, and that there needs to be better management of expectations thereof. However it is also critical that the lack of a "vanilla" or "one size fits all" solution should not be allowed to excuse the obvious lack of quality and ease of use in fundamental functionality. Too much energy is spent making new stuff. Too little on making the existing stuff any good.

Re: First Pages To Create In Your Private Wiki Notebook - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 16:31:10 -0000

This is an excellent idea. Gets past the "uh, so yeah, I've got one, what do I do now".

Re: Picking A Wiki For Your Private Notebook - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 16:29:47 -0000

I'm a huge fan of visible backlinks but got lots of stick for trying to make it a default in Socialtext. Was considered too much noise. They are available in TiddlySpace as an option (you can load them from a URI) but again: noise etc.

I think this feeling that they are noise come from a misinterpretation of why the wiki page exists at all. For many people they are merely a reference in isolation, not a member of a network. It's the network that matters.

Re: Why Use A Wiki For Your Notebook - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 16:22:24 -0000

"I'll hash that out more later in explaining how to Notice Patterns And Garden Your Private Wiki Notebook." is the most important/interesting part so you might want to give a teaser here?

Re: How Will Keeping A Notebook Help You Hack Your Life - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 16:18:09 -0000

"spent lots of" => spend lots of

The remind yourself of the big picture is a major payoff for what some people will see as the hard work.

Re: Private Wiki - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 16:03:39 -0000

I want my _private_ wiki to be able to see or be aware of twins in a semi-global context. My own pages wouldn't be visible unless I flipped them to being visible in that context.

Re: Wiki Types - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 15:53:32 -0000

Under-12, 12-?, ?-beyond seem like natural breaks. My own most value has been since in the first.

Re: Private Wiki - WebSeitz/wiki

cdent — Mon, 07 Oct 2013 15:51:49 -0000

I think a thing I would want in a PrivateWiki is a way for it to be a layer within/on/around other wikis so it can have TwinPages with other people because being able to compare and contrast with other thinkers is very useful (to me).

Re: Too long? Read anyway.

cdent — Tue, 26 Feb 2013 14:36:08 -0000

Very well said and very worth saying. You've articulated the motivation of pretty much my entire career: encourage thoughtful composition. Thanks for that because it can be easy to get a bit disoriented.

Re: The Chaos Collective: Where the Internet Went Wrong, and How We Can Fix It

cdent — Wed, 26 Sep 2012 17:40:25 -0000

I've been thinking about a similar concept called Internet Bank of Content for a while, first time with that name here: http://cdent.tumblr.com/pos... and then as a paper here http://iboc.tiddlyspace.com/