mastodon.ar.al is one of the many independent Mastodon servers you can use to participate in the fediverse.
This is my personal fediverse server.

Administered by:

Server stats:

1
active users

#queues

0 posts0 participants0 posts today
Continued thread

What’s interesting is that I seem to see this behaviour whenever there is a rush of new people onto the fediverse.

So I’m wondering how much is a bottleneck on my instance’s side (clearly, there is one) and how much is a general slowdown of message delivery/handling across instances (perhaps especially the major ones).

Replied in thread

With apologies to all the amazing system administrators out there for all the mistakes and gross simplifications!

References:
Mastodon: Scaling up your server >>> docs.joinmastodon.org/admin/sc
Scaling Mastodon in the Face of an Exodus, Leonora Tindall 2022/11/10 >>> nora.codes/post/scaling-mastod

docs.joinmastodon.orgScaling up your server - Mastodon documentation
Replied in thread

They’ll get it. They’ll analyze and improve processes. They will add more processing and storage. Once the queues start getting smaller then they have it licked! They can go to bed knowing that in the morning they will have caught up. Unless Elon does something crazy like let the former guy back on the birdsite and…

So if you can afford it, give some money to your instance. They need it now, not in…14hr.
9/9 more>>>

Replied in thread

They do this with more computers. Most of these larger systems use datacenters like you’ve seen in the movies. A big instance probably does not own the machines and is leasing the computing and storage on a monthly basis from a datacenter company. To add capacity they lease more stuff. Then they need to make it all work, again at sea in a storm with another huge wave bearing down. So you get the picture.
8/9 more>>>

Continued thread

In this situation (now it’s 13h) it‘s all hands on deck. The ship’s taking on water. Not foundering…yet, but got to get the pumps working! Normally, things might drop off at night and the system could recover. But, when the perfect storm hits like today and yesterday, the water just continues to rise, and they need bigger pumps. What’s more they need to get them working in the middle of the storm.
7/9 more>>>

Continued thread

That’s what the lower priority queue jobs do. If the server is overloaded those jobs don’t get done and the Federated timeline falls behind. But, the really important stuff like publishing your posts and sending them out do get done. If you start getting nasty 502 errors then those jobs are not getting done either. It’s all about processing power and the amount of work that needs to get done.
6/9 more>>>

Continued thread

Still there?… Mastodon uses different queues for tasks of varying importance. The most important queues get taken care of first. It’s like boarding group 1. Well, back in boarding group 4 and 5 are the jobs that fetch information from the other servers and mange # searches.

Wait what?... fetching information from other servers? OK…
4/9 more>>>

Continued thread

Every job has to line up at the back of the queue and the server services the jobs one after another. When things are working right the server has plenty of time to handle each job and the queues are very short. Everyone is happy.

But when there is too much work to do and not enough server power, the queues can back up. That is what is happening to your Federated timeline.
3/9 more>>>

Continued thread

If this is happening to you, go ahead and click on the world icon on the right. This shows you your Federated feed. Note that this is also running behind. In our case now 12hr behind. This is because there is too much work for the server to do and something has to give.

So hang on here… Mastodon has things called queues which are... queues for handling jobs the server needs to do.
2/9 more>>>

Repost From Nov 20 for #NewUsers

Wait! What's wrong with my bleeping # searches?

Maybe today you tried a search and there was something curious about the results. On this server almost every post returned was 5hr old, then it grew to 10hr, and now it is 11hr. What’s wrong?

Well, first thing is, Mastodon is not broken, but your instance is struggling.

1/9 Expand this post for a kind of techie description>>>

#Enterprise #Message #Queues require a different mindset than batchwise #processing. With an #EMQ one gets nudged to handle 1 message at a time. It needs #software #languages that have quick start-up times. Historically, #Java has a bad rep for that.

With #Java11 and IBM
's #OJ9 the cold-start speed of a program can be greatly increased (vs Oracle
's #Hotspot) to the point where the program no longer needs to run forever... which is why we can use it on #AwsLambda with #Kafka.