@polotek@social.polotek.net avatar

polotek

@polotek@social.polotek.net

Web developer, movie buff, and pretty much the best guy you know. Married to
@operaqueenie

This profile is from a federated server and may be incomplete. Browse more on the original instance.

polotek, to random
@polotek@social.polotek.net avatar

The main reason that productivity metrics for software development aren't useful is that we don't really have any standards for determining what outcomes we are shooting for. Nobody knows what "good" is. So we don't know what we're measuring for.
https://hachyderm.io/@jenniferplusplus/112010057412867182

polotek,
@polotek@social.polotek.net avatar

It goes the other way too. There are lots of engineering teams that are objectively not excellent, but the company is thriving. So many other factors play into company success. Bad products win all the time. So what's the point of a conversation with that team about how to do better? Money is a better metric than "productivity" 100% of the time.

polotek,
@polotek@social.polotek.net avatar

@flowchainsenseisocial rudeness appalls me.

polotek,
@polotek@social.polotek.net avatar

I know I tend to gripe and talk shit a lot. But this is something I spent a lot of time working on as a manager. Everybody wants some kind of system to know whether software engineering teams are doing a good job. But I found it pretty impossible to get people to align on what qualifies as successful outcomes. We have not succeeded. Instead, that space is filled with a lot of feelings and bad intuitions.

polotek,
@polotek@social.polotek.net avatar

This dynamic plays out in many different directions. Nobody is really happy with it. Sometimes software teams are kicking ass, but it doesn't matter because they're not able to hit the goals that the business really needs. This often looks like arguing that the goals are "unreasonable". That may or may not be the case. But it's irreconcilable either way. If the business isn't succeeding, then the team isn't succeeding.

polotek, to random
@polotek@social.polotek.net avatar

Last week was my first time really looking into what settings are available to admins who run their own instance. I was a bit surprised to find there isn't much there. At least through masto.host. I don't think they hide settings though. Only some of the deeper technical stuff.

polotek,
@polotek@social.polotek.net avatar

Here's the snapshot of my sidekiq status again. The take away here is that there was no huge spike this morning. This is normal usage. It still produced upwards of 15K jobs that need to be completed. That's just doesn't feel like a reasonable design for this kind of system.

polotek,
@polotek@social.polotek.net avatar

Not weird at all. We have been trained out of wanting to understand and control our own experience. But it's actually very cool and empowering.
https://mastodon.social/@Palleas/111998749838523835

polotek,
@polotek@social.polotek.net avatar

FYI @mastohost

I'm once again experiencing very slow read and write times since I dropped back to the lowest paid tier. Please let me know if there's something you can do.

polotek,
@polotek@social.polotek.net avatar

@mastohost I'm using @phanpy. But keep in mind this problem has persisted across multiple clients I have tried.

polotek, to random
@polotek@social.polotek.net avatar

I'm gonna spend the next hour or 2 learning a little more about performance in mastodon. I haven't looked at since this thread earlier in the week.
https://social.polotek.net/@polotek/111966879667140469

polotek,
@polotek@social.polotek.net avatar

Unfortunately, I didn't bother to look at this view when things were really slow. Especially when my threads were getting more than usual attention (because I was starting shit with fediverse nerds). But when we look at the 3 month history of jobs processed, we see big spikes. The all time peak of jobs was earlier this week and it shows 145,000 jobs. That feels like a lot of work to do for an account that is not very big.

polotek,
@polotek@social.polotek.net avatar

@thisismissem yeah I'm familiar with pghero. I appreciate your input so far. But you can dial up the technical dial on me bit.

polotek,
@polotek@social.polotek.net avatar

Here's the status view for sidekiq. This is the nerdy stuff. A few things I notice right off the bat. Sidekiq only has 4 processing threads. The tier that I upgraded to for masto.host is supposed to have 8 processing threads. But according to their FAQ, that's split between "federation jobs" "concurrent short-lived HTTP". So I think 4 is expected here. According to folks earlier this week, Sidekiq performance is directly correlated with what kind of performance you see in the UI.

polotek,
@polotek@social.polotek.net avatar

Here's the settings page with stuff @mikey was referring to. We don't care about deleting archives right now. The important values are empty by default.

I'm gonna set Media cache retention to 30 days. Content cache retention is more interesting. It says "All posts and boosts from other servers". Which means not my own posts. But I'm gonna do some double checking on this. Leaving it empty for now.

I'm pretty sure none of this is related to the performance issues I experience though.

polotek,
@polotek@social.polotek.net avatar

@thisismissem maybe I'm misunderstanding. If I run pghero, it still has to connect to the db hosted by masto right?

polotek,
@polotek@social.polotek.net avatar

Are there any other performance-related views on the mastodon Admin side that I should check out?

FWIW, this is the stuff the mastodon admin on your instance can see. It's not a lot tbh. If they're running their own servers, they may have more tools that they've set up themselves. But it's still not clear how to respond if users are experiencing poor performance.

polotek,
@polotek@social.polotek.net avatar

@thisismissem how would that work? Would they provide you a secure tunnel? That feels dicey.

polotek,
@polotek@social.polotek.net avatar

I consider myself a pretty heavy user. (Maybe you noticed). So this isn't an indicator of the value of the low tier for an "average" user. But something to keep in mind.

polotek,
@polotek@social.polotek.net avatar

To be honest, I've had a much better experience since I boosted my masto.host plan. I wouldn't say things are "fast" by any means, but it brought the average performance back within a tolerable range. But again, I don't think I'm gonna continue paying $19/mo. So I wanna see if there are other ways to make things better while dropping back down to $6/mo tier.
https://social.polotek.net/@polotek/111967626833457301

polotek,
@polotek@social.polotek.net avatar

In the admin settings on my instance, I can see certain views related to performance. Folks told me about sidekiq earlier this week. I'm digging onto that now. But there's also a tab for pghero. It's blocked by masto.host. This makes sense for an instance managed by a company. But if you wanted to run your own instance on a server, you could see database stats. That's cool.

polotek,
@polotek@social.polotek.net avatar

This is fair. It's not clear if the caching or whatever is included when calculating whether you've hit your plan limits. If they are, then it's still relevant in terms of growth.

Also, I assume you mean deleting other people's posts that were federated to my instance. That's fine. I never delete my own posts. It's a rule I built on twitter, and plan to maintain on all social platforms. It keeps me accountable.

https://friendsofdesoto.social/@mikey/111983640216541256

polotek,
@polotek@social.polotek.net avatar

Oh, a couple of corrections.

The peak of jobs was last week, not this week.

And I haven't really been using mastodon for 8 months. I created the account 8 months ago, but didn't really use it. I only really started using it 3 months or so ago. You can see that in the chart. That makes the storage growth numbers look worse tbh.

polotek, to random
@polotek@social.polotek.net avatar

I still maintain that the at protocol means that bluesky should be included in the definition of "the fediverse". I know it will take people some time to come around on that. And certainly bsky still has some promises to deliver on. But the actual fediverse will be so much messier than what people have in their heads right now. The lines will all be blurry and gray. By design.

polotek,
@polotek@social.polotek.net avatar

@Gargron yes I understand the perspective. I still disagree with it. Explaining it again doesn't change anything.

polotek,
@polotek@social.polotek.net avatar

Me, clearly and explicitly: "I have this idea. It's unpopular today, so I'm sure it'll take people a while to come around on this idea..."

Other people, immediately: "No! I absolutely disagree with you at this point in time!"

Me: "Yes, I know. That's why I... you know what, it's fine. Thanks for stopping by."

polotek, to random
@polotek@social.polotek.net avatar

One of the big problems at the center of open source projects that become real products. There are people who like to talk about code more than the humans who are experiencing the code. That's fine. But those should not be the same people you reach when you're looking for product support.
https://social.polotek.net/@polotek/111971552688471590

polotek,
@polotek@social.polotek.net avatar

@WTL the Linux community is the absolute worst for this.

polotek, to random
@polotek@social.polotek.net avatar

Okay. I guess I'm gonna make this account a little more annoying. I'm gonna spend a little time each week exploring my mastodon experience and talking about how to make it better.

I say "annoying" because I'd rather be interesting than just useful. And I think this kind of content is the opposite. Potentially useful but overall not the most interesting. (It's okay if you disagree. We don't have to talk about it)
https://social.polotek.net/@polotek/111966798276024291

polotek,
@polotek@social.polotek.net avatar

Posting is also super slow at times. The above post with the two images took around 30 seconds. That is unsustainable.

polotek,
@polotek@social.polotek.net avatar

@mastohost @alec @jacob as I stated above, this happens with multiple clients.

Are there any upgrades to memory with the pricing tiers? Only processing threads and storage are mentioned explicitly. https://social.polotek.net/@polotek/111966916310343916

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • Leos
  • ngwrru68w68
  • InstantRegret
  • thenastyranch
  • magazineikmin
  • GTA5RPClips
  • rosin
  • osvaldo12
  • tacticalgear
  • Youngstown
  • slotface
  • khanakhh
  • kavyap
  • DreamBathrooms
  • JUstTest
  • modclub
  • everett
  • provamag3
  • cubers
  • cisconetworking
  • ethstaker
  • Durango
  • mdbf
  • anitta
  • megavids
  • normalnudes
  • tester
  • lostlight
  • All magazines
    •