|
9 posts total
 I'm looking for a little help working through some database scaling issues with #WriteFreely. On our Write.as instance, we're up to 4.25 million rows in our `posts` table, and it's causing problems for some individual blogs with 9,000+ posts. I have an idea of what some bottleneck queries are. But wondering if there are easy db optimizations we might make (indexes, etc.), and if anyone can help identify what they might be. Boosts appreciated! Writing.Exchange registrations are completely open for now, while people are checking out Mastodon again  @matt I hope you have enough people to keep eyes on reports if they come in! But also Yey. 😅 @matt That's great. I am sorry I've been absent for so long. I will try again to do my part to help create a sense of community for writers. @matt hi, matt. While you’re here I was wondering if you would have any idea why I’m not able to access an account that I have on another mastodon server (@linernotes.club, same handle). Also, what would you recommend in terms of say an annual donation amount? (As you can tell, I’m a recent arrival) Ugh, I'm locked out of a seldom-used @Mastodon account, on an instance I host that only I ever use, because I never set up email on that server. So I can't get the security code since I "appear to not have logged in from this device before". Is there no admin option to override the login block, @Gargron?  @matt There's indeed none (which is not good, I know). If you set your current_sign_in_at to a current datetime from the rails console it won't ask you for the code. This is very odd... One of my #Mastodon instances says for an entire #WriteFreely domain: "delivery is not available," while another Mastodon instance says the same domain can be delivered to. 🤔 Following doesn't work from the "domain undeliverable" Masto instance. I don't think there's any moderation-related reason this domain would be marked this way. So what would cause this? cc @Gargron @matt Delivery unavailable = no more delivery attempts after 7 consecutive days of delivery failures As I wrote about recently on @matt (https://write.as/matt/falling-short-in-2020), I'm spending my spare time thinking about a new private communication tool that helps me keep in touch with people better. Right now, it's looking like it'll be a new email client. Biggest ideas are: - An "inbox" interface that's more about people than messages Only got to catch the end of Evan Prodromou's #APConf2020 Q&A session, but I really appreciated the point he made about how we might grow the fediverse by only having certain interactions here (instead of cross-posting to birdsite, for example).  |
Gregory's Server
@matt if you haven't already, turn on slow query logging, it will expose the problem queries and you can tune indexes from there. You didn't mention the DB engine you are using, but MySQL can be an absolute nightmare to do this, I do it all the time.
@matt Unfortuantely I know nothing about your stack or your code (other than it being MySQL-backed). However, here's the general troubleshooting lists of things that make a database-backed web application slow:
1. Too many queries: app is pulling lots of things from the DB constantly that could be cached, or doing joins in the app code
2. Unoptimized requests: some DB requests need better queries/better data structure/indexes
...