Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Ask HN: Is HN slow today?
156 points by brundolf on Jan 4, 2021 | hide | past | favorite | 76 comments
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?


Yes, 'Who Is Hiring' days strain our server in general and there are a lot of other things going today, like Slack being down and the holidays being over. (Edit: and one mammoth 500-comment subthread that I've since split in two.)

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.


This is the first time I've ever had a slow experience with HN, so there is no need to be sorry. A damn fast website can have a few slow downs once in a while in my mind. HN is the opposite of slow software, and even at its slower moments, it still beats the majority of the modern web (looking at you, Reddit redesign).


It's all relative. I find it too damn slow.


Time to RIIR? (... Sorry, I couldn’t resist)



I don't want to come across as contentious but will. you. please. just. use. a. real. database.

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're very obnoxious


Please don't post like that to HN, no matter how obnoxious someone is or you feel they are. It only makes things worse.

https://news.ycombinator.com/newsguidelines.html


Maybe. Many people here are. It's still a valid criticism.


No need for the apology I think. You guys do a great job keeping this place simple and fast for us.


I hope you don’t mind me asking but what’s the purpose of having a comment in the top to direct to different pages. There is already a more button on the bottom which makes the top comment kind of redundant.

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'm super curious about HN metrics like number of visitors, RPS at peak hours, details about the db, is it still flat files?


Does that affect HN replies? Or does it have other issues?

I have not received any notification mails in the last days


It's still pretty fast for me. I wouldn't have noticed if not for this post.

You guys rock!


> Yes, 'Who Is Hiring' days strain our server in general and there are a lot of other things going today, like Slack being down and the holidays being over. (Edit: and one mammoth 500-comment subthread that I've since split in two.)

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).


> A bit embarrassing HN should not run as if if it was operated out of one person's basement, budget, or expertise.

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.


On the contrary... Running a site like this with the hot new rube goldberg stack and drastically overpriced VMs would be embarrassing.

I find the simplicity here quite comforting.


> 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.

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.


It's single-threaded, isn't it? If so I'm sure the hardware is fine and not the bottleneck.


It's not just you...

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..." "


> 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...

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).


HN caches pages for logged out users (since they'll obviously be the same for everyone).


Lol. Circle-K indeed.

I also had some craziness on LinkedIn earlier. I’m not saying it is related, just correlated maybe.


>I also had some craziness on LinkedIn earlier.

Sounds like a normal day on LinkedIn


What if SkinWalker Ranch had become well-known before Bill and Ted?

In any case, I thought it was just me ;)


huh?


The log-in page before logging in is also slow to load.


I think this is normal, when ever a crucial Internet services or Network is down. That could be Azure, GCP, AWS, Verisign, Cloudflare, DNS attack etc.

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 haven't seen any issues with AWS. Slack seemed to be having it's own issues. Notion is having an outage as well. But otherwise, it's not widespread and none of the AWS services my group runs are having any issues. Sometimes these are coordinated DDOS attacks, but I can't imagine who would target Slack, Notion and Hacker News.


I don't believe anybody's suggesting Slack and HN are being targeted, more that Slack being down means HN is being hugged to death by actual users.


ksec didn't imply hn is a DDoS target. They pointed out that hn is usually rather slow when there are largescale infrastructure issues, because people come debate about them here.

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.


When google went down recently, HN went completely offline for a while. I suppose this is a side effect of lying status pages.


Slack seems to be back. Will be interesting to see how long slowness on HN persists.


And HN is back to normal as well :)


Not sure if this is related, but a traceroute is showing a lot of hops between my router and HN.

Me -> My ISP -> (Cogent: AMS -> PAR -> BIO -> PAR -> BIO -> DCA -> JFK -> DCA -> ATL -> DCA -> IAH -> ELP -> PHX -> SAN) -> HN


That seems a bit excessive. I have me -> My ISP -> (Level3: BOS, SAN) -> HN


Yeah... But now HN is fast again, FWIW.


Yes, same here. I'm wondering if it is the post holiday rush? Everyone back to work and getting on HN (and Slack and YouTube and everything else that seems slow today)?


Alternatively: HN is everyone’s Slack fallback and we’re just hard at work here.


Would also throw in the pandemic factor.

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.


My guess is theres some sort of yet to be identified global issue. Seems Comcast has been spotty in parts of the US, slack is down, youtube was briefly inaccessible for me, etc.


I'm definitely betting on the slack outage factoring in here. HN is my go-to to see if I'm the only one having an issue, and slack being down is distracting enough, and disruptive enough, that I'm definitely guilty of browsing HN more than I normally would, even for the first day back after a holiday break.


The same question was asked when Google was down. Apparently this leads to a surge of traffic to HN because everyone wants to know what’s up and this causes the slowdown. The recommended ‘solution’ is to log out so that HN can serve you with the static version of the site.


I think there may also be a general surge for HN since it's the "back to work" day after the holidays. Perhaps some largish portion of HN users didn't use it much over the holidays and are catching up on reading.


I've had some applications down today. The initial reports are date related issues with the New Year. I suspect that's incorrect given it's Jan 4, but an anecdote nonetheless.


Markedly slower than usual here - South Wales, UK - and has been for a couple of hours.

Edited to add: as per comment by @peter_d_sherman all slowness disappears if using the site without being logged in.


Notion is sending out emails that their outage/slowdown is due to AWS outages but AWS status board is green and I haven't seen slack or anyone else claiming aws issues.

> 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.


I use Hacker News as a quick benchmark for "how fast is the Internet here?" because the speed is so consistently good. Typically slow Hacker News load = slow Internet.


For me it was fine an hour or so ago when someone mentioned it in the Slack discussion, but it has steadily gotten more laggy as time goes on. It's sluggish now.


Yes. Page loads about 10 seconds, posts about 20 seconds.


I haven't noticed it being slow but it is better then failing completely at rendering the page (when it's under a heavy load).


Noticed this as well.

Edit: 7+ seconds to add a comment.


it isnt just HN. Slack is down, Notion (who is already slow) is somehow even slower today.


Notion emailed me saying it was because of AWS experiencing problems but no update to aws status boards yet...


Issues like this might be DNS related...

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


Plenty snappy for me, although I wasn't on much this morning.


Yeah, it took 5+ seconds to load this page.

Edit: And 5 more to edit this comment


Yup; consistently 7, 8, 9s to load every page.

EDIT: 19s for that comment :/


Oof.


And now it's fast again.


I saw some errors on HN yesterday, for what it’s worth.


Commenting seems gone now, too, links to search wtf?


I see yours, and am able to reply. ...and edit... a second or so of lag perhaps. via Xfinity/SF


Seems like I posted a comment to a withdrawn story.


Yea the slackers are not slacking today lol


Does it have anything to do with what's going on now? https://downdetector.com/


What is going on, exactly? Are there any theories?


Though to pinpoint an exact issue, but i'd say that the return to work from a holiday weekend exacerbated whatever problem exists.


Jikes. AWS, Google and Azure all spiking.


Doesn't HN run on a single computer under some desk? I heard this one time but never verified. Maybe it's time for a proper cloud deployment?


It's on a single computer, with a second as failover, but it's in a colo center, not under a desk. Well, I suppose it could be under a desk.


I remember something about the database being a flat file / flat files. Is that (still) true? How are the two machines being synchronized?


You can always run it over NFS.

I'd be curious to peek at HN Source code. I recall there was a version posted online but it's several years old.


Maybe dang deploys to HN using ftp ha ha ha


Deploy what? Hasn't HN been the same for years?


There are new features/changes added often, many are documented here (not with dates, but you can get an idea of when things change based on the commit history): https://github.com/minimaxir/hacker-news-undocumented

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).




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: