Ask HN: Is HN slow today?
What's normally the snappiest site on the web has been consistently taking ~5 seconds for each page load today. Extra traffic due to Slack being down, perhaps?
Sorry. I hate slow software.
Edit: for the time being, I've lowered the number of comments that get rendered per page. That's helping some. It means you'll all have to click more More links at the bottom of threads to see all the comments. One day this will all be better.
A flat-file architecture only scales so far. This is probably the only forum on the internet where a thread passing 500 comments is an actual problem - and remember, you're planning on getting rid of pagination, which means no theoretical limiter on request sizes. In a few years when 500+ comments threads become the norm among another million users, disk I/O is going to be more and more of an issue.
Racket has a perfectly good database library, just kick the problem upstairs and be done with it.
You guys rock!
You have access to both highly qualified people (that read HN and would most likely be glad to help you engineer a system (paid or unpaid) that would not be strained by usage) AND access to money if that was needed to improve things (and pay for a solution).
Does HN still run on a server colocated in a data center?
Is there a budget that you need to adhere to? YC certainly makes money and can afford to pay for a better setup.
A bit embarrassing HN should not run as if if it was operated out of one person's basement, budget, or expertise. (And it's not 1996 either it's 2021 and HN does not even have anything but text).
To me, this is a hopeful message. It is still possible to run a website out of one's basement and have it become a staple of internet culture.
Why is a colocated server considered a bad setup? I think that actually getting the resources you pay for (instead of being on an overprovisioned VM host) and not being scammed on network data transfer fees is a great setup.
I find the simplicity here quite comforting.
Also a tip for others when HN is slow or almost down (like when Google was down): you can open a private tab and navigate HN from there. Since you’re not logged in, you’re served a copy from cloud flare. It’s usually out of date by 1min or so but good enough for reading in my opinion.
I have not received any notification mails in the last days
I posted the following in the "Slack Down Completely" article (before this article was posted):
>"For some reason, today, HN seems exceedingly exceedingly slow (to me) after logging in...
Without being logged in, things are as fast as they usually are -- but post log-in, SLOWWWER THAN MOLASSESS...
I tried this several times; why this is, I can only wonder...
To quote Bill and Ted... "Strange things are afoot at the Circle-K..." "
I also had some craziness on LinkedIn earlier. I’m not saying it is related, just correlated maybe.
Sounds like a normal day on LinkedIn
> I tried this several times; why this is, I can only wonder...
That is to be expected. Non-logged-in pages can be cached aggressively server side. When you are logged in much more has to be done dynamically.
The fact that non-logged-in pages are still loading fast is testament to the fact that the HN devs have done their job properly with implementing caching.
In the bad old days it would be slow for everyone whether logged in or not. In general I mean. I have only been on HN since 2009 and have been logged in (other accounts in the past, changed name) pretty much always so I don't know when HN got good at caching. I would kind of sort of expect that HN was actually like that from the beginning, given that pg came from having used Lisp in production at Viaweb and they got a lot of experience there early at how to do things fast (in terms of time to implement) and presumably proper (since they were able to compete with others).
In any case, I thought it was just me ;)
A huge number of people comes to HN to check status. To the point HN report network / services condition earlier than any mainstream or social media. (Including their own Status Update pages ) Last time it was Google ( Storage Limit )
This time it is Slack.
I've noticed it many times in the past year.
E.g. when Google services had the big outage, or when AWS US-EAST-1 was down, or when Facebook SDK broke all the iOS apps, etc.
Me -> My ISP -> (Cogent: AMS -> PAR -> BIO -> PAR -> BIO -> DCA -> JFK -> DCA -> ATL -> DCA -> IAH -> ELP -> PHX -> SAN) -> HN
Whatever post holiday peak back-to-business load was before the pandemic probably just got a lot higher given the amount of people now working remotely opposed to before the pandemic.
Edited to add: as per comment by @peter_d_sherman all slowness disappears if using the site without being logged in.
> For the last couple hours, Notion has been down. This is related to some difficulties Amazon Web Services is experiencing today. As of right now, our engineering team is all hands on deck working to restore service. We're confident these issues will be resolved soon, but in the meantime I'm so sorry for any problems or inconvenience this has caused. I know this is not ideal as everyone returns from break, and we'll be thoroughly investigating how to prevent outages like this in the future.
Edit: 7+ seconds to add a comment.
If anyone needs to check the state of their DNS on a bunch of servers at once check out dug: https://git.kaijucode.com/matt/dug
Edit: And 5 more to edit this comment
EDIT: 19s for that comment :/
The most recent change I remember was when Twitter and GitHub posts were split based on the user the link belonged to (i.e. GitHub.com/dang is different than github.com/torvalds now).
I'd be curious to peek at HN Source code. I recall there was a version posted online but it's several years old.