Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Peter McGowan
    @pemcg
    it looks pretty cool

    I have a few different sized vcsims, PM me if you want to use one

    Thanks

    Dennis Metzger
    @dmetzger57
    Thanks for that pointer @pemcg
    Peter McGowan
    @pemcg
    np @dmetzger57
    Keenan Brock
    @kbrock
    I had tried the go simulator a while back and it was missing a lot of the data
    But the striking thing about the go simulator is that it seems to be part of a rewrite of the engine and isn’t just some bogus simulator.
    The cvs I’m is a real environment but is just so big
    Ugh. The vcsim is bloated but uses the real code, which is good. Think it is deprecated
    I need to try out the go one again. It had so much promise and good VMware backing
    Keenan Brock
    @kbrock
    ^ ignore. Very excited to hear your experiences with the new go sim.
    Peter McGowan
    @pemcg

    ^ ignore. Very excited to hear your experiences with the new go sim.

    Well I had no experience with the old one, so seeing 2 emulated VMs is exciting for me :smile:

    Adam Grare
    @agrare
    yeah depends on what you want to test, the nice thing about the old vcsim is that it is a real vcenter so it will exhibit the same behavior as you'll see in a prod env
    so if you just want to see "what does 1000 vms look like" I'm sure govcsim is fine, if you want to see how your code will work on a real vcenter with 1000 vms I don't thing govcsim will do that
    Peter McGowan
    @pemcg
    ok that’s useful to know, thanks. For now I just want quantity of VMs, hosts, clusters etc to get an idea of timings and memory usage
    Keenan Brock
    @kbrock
    the basics in the go were working even a while back. I was concentrating on metrics at the time and the go api just wasn't there yet
    it sounded like the go simulator was possibly a front end for some future vcenter work - wonder if they are thinking about doing a major revamp - at least of the soap front end. Which makes sense. Quite a few vendors rewrite the web service layer in go. well in the ruby world at least.
    Daniel Berger
    @djberg96
    anyone else see a local failure in spec/tasks/support/gem_base64_spec.rb:73 in manageiq-performance?
    ruby 2.6.6 on catalina if that matters
    Keenan Brock
    @kbrock
    huh - just saw https://github.com/Shopify/app_profiler - interesting
    Jason Frey
    @Fryguy
    cc @gtanzillo @jrafanie @bdunne maybe this is part of why we see things are "slow" on boot in docker
    I believe that the cost of two syscalls (open(2) only after successful access(2)) would, in our case, at least because we would shave-off something like 1,900,000×90µsec (2.85 minutes) from the three minute boot time for our application.
    Brandon Dunne
    @bdunne
    Interesting. In a well performing environment this isn't noticeable, but in the slow :fire: environments this could make a huge difference
    Jason Frey
    @Fryguy
    yeah
    and we fall in that bucket of "huge application with a lot of load paths"
    Brandon Dunne
    @bdunne
    yep
    Jason Frey
    @Fryguy
    wonder if we can measure those syscalls in the way he did
    also it seems zeitwerk helps a lot because it avoids that file search that require does
    Nick LaMuro
    @NickLaMuro
    errr... I thought part of that thread made it seem like it did not help
    Jason Frey
    @Fryguy
    oh I missed that...
    Twitter threads are hard to follow haha
    Nick LaMuro
    @NickLaMuro
    agreed. Took me a minute to figure out the original context from that tweet, so I ended up having to look up to get some context. Still really interesting, and yes, relevant
    Joe Rafaniello
    @jrafanie
    I was able to get the gist much easier by reading from here down: https://github.com/docker/roadmap/issues/7#issuecomment-757710368
    Jason Frey
    @Fryguy
    @kbrock thought you might find this interesting https://twitter.com/josh_cheek/status/1361472100012347395?s=19
    NickLaMuro @NickLaMuro is trying to figure out if that is executing the loop once... or every time the query is executed... (me thinks the latter)
    Keenan Brock
    @kbrock
    @Fryguy I love how they stored the hierarchy in an array - really want to do that
    Jason Frey
    @Fryguy
    generated it, but yeah
    i think the storage is just parent_id
    Keenan Brock
    @kbrock
    for hierarchical queries - yes, just parent_id
    I find array_arg and a couple of the "group by" type array creation queries confusing
    understandable, but it requires a little extra groking for me
    Keenan Brock
    @kbrock

    whoa, I totally missed this: https://jemma.dev/blog/irb-measure /via ruby weekly 1/21

    ruby 3.0 has measure :stackprof or timing for built in metrics.

    Nick LaMuro
    @NickLaMuro
    that is neat, though still not sure if stackprof is bundled with ruby by default
    I have longed thought that it SHOULD BE, but this might be one of those scenarios where Aaron added that as a default measurement, but didn't included it with the ruby lang by default
    Keenan Brock
    @kbrock
    Remember tech empower benchmark talking about db drivers in some programming languages multi-plexiglass requests or something. Still trying to track down.
    BUT I did stumble upon pgbouncer (reminds me of dbslayer from NYtimes 10+ yes ago?) wish rails did not have so many begin/end transactions and were statement centric but I digress. https://www.percona.com/blog/2018/06/27/scaling-postgresql-with-pgbouncer-you-may-need-a-connection-pooler-sooner-than-you-expect/
    We create so many connections that a pooler may make sense? Dunno
    Lol. Plexiglass. Thnx Siri
    Keenan Brock
    @kbrock
    Aah. Pipelining. https://www.techempower.com/blog/2018/10/30/framework-benchmarks-round-17/
    I’d imagine some programming languages have this while others do not. Also, some forms of pooling would probably work better for this than others
    We have so few clients and our number of queries are high per client. Also the query complexity is high per call. Our app is probably not a good contender.
    Does seem like a concept of interest
    Keenan Brock
    @kbrock
    Aah. Maybe this is the adequate record as be prepared statement stuff. But that was a long time ago
    Think this is moot because rails loves to dynamically create sql (there are always tons of ? Per request that are dynamically altered - trashing query reuse) wonder if we could use ANY instead of IN to pass a single array instead of passing a variable number of parameters.
    Yea. This is probably too low level and fighting too much with the nature of rails
    Keenan Brock
    @kbrock
    And in conclusion. My question of “how hard is it to just add this pipelining to the driver?” Has an easyish answer. You’d probably need to rewrite the driver to not use libpg and need to write some code that leverages this - that is a bit much. And ruby is slow. So it looks like libpg needs to be extended to handle a better c api for pipelining. Which seems like someone would have already done this... guess that is my next step before giving up