Tell HN: Server Status
101–110 of 124 posts
Re: Tell HN: Server Status
#102Earlier quoted context omitted.
Not to discount the value of HackerNews, however there is great benefit to YCombinator being associated with it; They could let it go down and never host it again but they would lose value.
Yes, but I don't think such shortages do them much harm.
Hard to quantify "harm".
But I will say that anything that is a habit, when broken, opens you up to the possibility of being exposed to a new habit "the addiction".[1]
Along those lines 1 day of downtime probably isn't going to shift attention much. But an extreme, 2 weeks, would certainly break some addictions as people would fork to a new (I don't know insert some french or latin word here!) So who is to say where that slippery slope of is? (Somewhere between 1 minute and 2 weeks).
[1] Reason that I have heard that Starbucks renovates restaurants at greater expense and keeps them open at least some hours. Because people's habits are mercurial.
Re: Tell HN: Server Status
#103Re: Tell HN: Server Status
#104Why on earth are you not using SSD's? The HN footprint can't be that large. The extra speed and reliability from a pair of SSD's has to far outweigh the costs.
Re: Tell HN: Server Status
#105Hi, I'm the CEO of http://www.clever-cloud.com/ and I'll be happy to help you on this, ping me on twitter : @waxzce
Can you prove that you are the CEO? I'm doubting between this being a bad prank from someone impersonating you, and you actually placing advertisements for your own company here.
Re: Tell HN: Server Status
#106Add a flash memory based (SSD) ZIL or L2ARC or both to the box. That'll help improve read/write performance. I believe the ZIL (ZFS intent log) is used to cache during writes, and the L2ARC is used during reads.
You might want to look into disabling atime, so that the pool isn't wasting energy keeping access times on files up to date. Not sure if this is relevant with the architecture of HN or not. This can be done with
zfs set atime=off srv/ycombinator
Finally, ZFS needs a LOT of memory to be a happy camper. Like 3-5GB of RAM per TB of storage.I actually think you'll probably have a lot of fun with ZFS tuning, if that's the problem with news.yc. FreeBSD's page is pretty detailed: https://wiki.freebsd.org/ZFSTuningGuide
Re: Tell HN: Server Status
#107Good you posted this, but it came a little late. After the first series of timeouts you could've posted an update so everybody knew what was going on. But hey, thanks for the update, this clears up a lot.
Re: Tell HN: Server Status
#108I've been reading this site regularly for almost 7 years. 6-Jan-2014 is the only downtime I remember, and it was really a very minor inconvenience. Sucks about the data loss though, always hard to own that when doing system administration. Thanks for the explanation.
Re: Tell HN: Server Status
#109Earlier quoted context omitted.
Why not Raid 60 with Btrfs? It'll tolerate two disk loss with pro-active parity protection via btrfs and be faster and provide you with more disk space.
Because BRTFS isn't even feature stable yet.
Re: Tell HN: Server Status
#110Earlier quoted context omitted.
Yes, but I don't think such shortages do them much harm.
"do them much harm." Hard to quantify "harm". But I will say that anything that is a habit, when broken, opens you up to the possibility of being exposed to a new habit "the addiction".[1] Along those lines 1 day of downtime probably isn't going to shift attention much. But an extreme, 2 weeks, would certainly break some addictions as people would fork to a new (I don't know insert some french or latin word here!) So…