Perl 6 - the future is here, just unevenly distributed

IRC log for #crimsonfu, 2016-05-04

crimsonfu - sysadmins who code

| Channels | #crimsonfu index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:38 ijstokes joined #crimsonfu
01:47 ilbot3 joined #crimsonfu
01:47 Topic for #crimsonfu is now http://crimsonfu.github.com - ConfiguRatIon Management of Systems Or Network kung FU | logs at http://irclog.perlgeek.de/crimsonfu/today
01:54 ijstokes joined #crimsonfu
02:12 ijstokes joined #crimsonfu
03:32 ijstokes joined #crimsonfu
06:59 Azgarech joined #crimsonfu
12:39 Azgarech joined #crimsonfu
13:17 skay joined #crimsonfu
13:39 Azgarech joined #crimsonfu
14:42 Azgarech joined #crimsonfu
16:13 Azgarech joined #crimsonfu
19:17 ijstokes joined #crimsonfu
19:17 ijstokes who has used openstack recently? fun or pain?
19:22 dotplus "yes"
19:22 dotplus what do you want to use it for?
19:25 dotplus longlived instances hosting production services? ephemeral instances spun up by a CI stack? a small isolated cluster of dev instance? do you have complex networking and/or storage requirements?
19:28 dotplus do you want to use more than just the minimal set of projects (auth/compute/network/image) or do you want to get fancy such as with db/hadoop as a service or other non-basic storage options?
19:34 bene the bluebox guys are down the street here in boston and i chat with them at meetups and the like
19:35 bene they are pretty persuasive on the "you should never, ever, ever run your own openstack cluster" sales pitch
19:42 dotplus I work on the Cisco team that does an Openstack product/service. my opinion (not necessarily my employer's) is that it depends what you want.
19:47 dotplus if your needs are simple and your outage/failure-tolerance is high, go ahead. If/when it becomes important or widely/heavily used or sophisticated or business critical, get some help. Help can come in various forms: experienced consultant/or employee(s), one of the many private/public/hybrid providers.
19:48 dotplus Just bear in mind that a) it's not a silver bullet b) done with appropriate expectations, it can be a hugely valuable tool
19:56 larsks ijstokes: oh, hey, I do a bunch of openstack! :)
19:56 larsks Prety much +++ what dotplus said there.
19:56 ijstokes hey larsks -- and I'm sure I can get an unbiased opinion
19:59 bene that's such a cryptic statement in an irc channel
19:59 larsks Yeah, I thought there was going to be some sort of follow up :)
19:59 bene heh
19:59 ijstokes Mirantis?
19:59 larsks They are a company that provides openstack support, yeah.
20:02 larsks ijstokes: still there? :)  I can't comment in particular on Mirantis, since being pretty much in the same line of work I haven't ever had to avail myself of their services.
20:02 larsks I hear they are good people.
20:02 ijstokes I've spoken with Mirantis on the phone.
20:02 larsks I think OpenStack is fun if you really like system administration work.  Otherwise, probably "pain", because maintaining openstack is going to be between you and what you actually want to be doing.
20:03 ijstokes That seems like the opposite of what it is supposed to do, isn't it?
20:04 larsks I don't think so, no.  It's just that maintaining a complex distributed system is Work, and it probably needs to be a full time job for somebody.
20:04 larsks Everybody else gets to have the fun.
20:04 ijstokes Our initial use case is to stop using Digital Ocean and instead leverage our in-house resources using OpenStack to allow our engineers to self-serve provision their VMs and VM clusters.
20:04 larsks That seems like a good use case.
20:04 larsks But you need to have somebody to replace all the digital ocean OPS staff who are currently making your life pleasant-but-expensive :)
20:04 pdurbin ijstokes: but digital ocean is only 5 bucks a month. I know because I have a little VM with them. :)
20:04 ijstokes But the *hope* is that there isn't a huge stack of additional sys admin overhead incurred from deploying OpenStack.
20:05 ijstokes Yeah, so the threshold for sys admin maintenance/support time on an in-house OpenStack cluster is pretty low.  Digital Ocean is crazy cheap.
20:05 larsks If you're not actually using a full time employee's worth of money annually on DO right now, you might just want to stay there.
20:06 dotplus a good use case especially if those VMs/clusters are ephemeral or shortlived.
20:06 ijstokes We always imagine they are short lived.
20:06 ijstokes And then (like today, when doing openssl housekeeping) we discover they aren't
20:06 larsks Set up something that kills off everything more than 24 hours old, regardless of what it is being used for.  that encourages flexible design :)
20:07 pdurbin ijstokes: are you engineers want CI, basically? If so, maybe bear can recommend a cloud CI vendor. :)
20:07 ijstokes But still, we'd need to have an aweful lot of $80/month 16 GB/160 GB/4 core machines sitting around to add up to a sys admin salary.
20:07 pdurbin your*
20:07 ijstokes pdurbin: all your base are belong to us
20:08 ijstokes We use a mix of CI systems, from Jenkins to Travis to CircleCI and since that wasn't enough we even built our own.
20:08 dotplus yeah == larsks. if DO is working for you and not costing a fortune, it might well be a good idea to stay with a public cloud provider.
20:10 pdurbin You just can't have enough CI.
20:10 ijstokes <sigh>
20:11 larsks ijstokes: if it meets your needs, something like kubernetes might be easier.  It's a little more all-in-one, as long as you can work with containers rather than actual vms.
20:11 dotplus if the only driver to in-house openstack is cost (and there are several other possible drivers that might make sense), then it's sort of easy to decide: in order to do openstack "right"-ish in house for a largish team, you'll need a >1 FTE @$125k/year (US tech-metro pricing)
20:12 bene + overhead + money for hardware and datacenter space
20:12 ijstokes dotplus if only $125k (salary only) was the only cost of a sys admin.
20:13 dotplus absolutely.
20:13 ijstokes benefits + employer taxes + overhad = 50% on top (maybe more?)
20:14 larsks + hardware + cooling + power...
20:14 dotplus I guess that did sound like I was saying that if your spend @DO is > $125k it's likely worth bringing it in house? No, I'm certainly not saying that. quite the opposite.
20:14 bene i think the more important point is the FTE part
20:15 bene title: openstack wrangler
20:15 bene "what do you do for work?"
20:15 bene "mom, let's not go through this again please."
20:17 pdurbin even mom has heard of the cloud
20:20 larsks pdurbin: I once tried to explain cloud computing to my seven year old daughter.  I explained that it was a way for people to share computers if they only needed them *some* of the time.
20:20 larsks pdurbin: She thought a bit and said: "Dad, that's the stupidest idea I ever heard."
20:21 bene Incompatible Time Sharing System
20:21 bene https://en.wikipedia.org/wiki/Incompatible_Timesharing_System
20:23 bene don't fall down the wikipedia hole there
20:24 pdurbin larsks: we just point at the router and say if the cloud is down you can't watch netflix on your ipad
20:24 bene i started reading about 6 bit character codes and realized that was probably not relevant to my task today
20:25 * bene . o O ( 6-bits + parity is perfect for 7 track magnetic tape! )
20:28 * larsks starts reading wikipedia about IBM 7 track...

| Channels | #crimsonfu index | Today | | Search | Google Search | Plain-Text | summary

crimsonfu - sysadmins who code