sweaner wrote:Ironically, Coffeegeek has been faster the last few days.

Perhaps you should spend more time there.

Dan, not even close!! Are you trying to give me a hint?

Oh bother.

General Error
SQL ERROR [ mysql4 ]

MySQL server has gone away [2006]

An SQL error occurred while fetching this page. Please contact the Board Administrator if this problem persists.

I've seen this error a few times in the last two days.

It is looking more and more like grid hosting = gridlock and that the more traditional approach may be necessary. If the site performance hasn't settled down by mid-next week, I'll pull the plug on this experiment and revert to a dedicated server with more memory. This time around I'll handle the DNS business more elegantly by keeping both servers up but pointing to the same database, so it won't matter if there's a propagation delay.

Eww, MySQL 4!

Dan, one EASY way to fix the DNS issue is to, at least 1.5 TTL (Time To Live) periods BEFORE the change (in other words, if you have a TTL of 86400 seconds [1 day] then, at least 1.5 days before REDUCE the TTL to 300 [300 seconds, 5 minutes]). The reason for this is to insure that DNS changes propagate out fairly quickly (and reversion BACK, if needed, happens quickly as well! ). THEN, 24-36 hours AFTER you're sure the change is "good" you can revert back to the standard TTL of 86400 seconds. This reduces DNS traffic back to a more "normal" level.... NOW, one "problem" with this is that Windows, in it's infinite wisdom, pays little to NO attention to the TTL values (I've got servers on our internal network that have NEVER updated their DNS entries, even 3 years later, for a system that had a 300s TTL SPECIFICALLY because the DNS entry pointed to 2 or more systems for a crude form of "load balancing"). Vista has FINALLY, apparently, done SOME towards that, but.... YMMV.

I did actually get a fairly good response for just a short period today... Went to a specific forum subject, which loaded in about 2-2.5 seconds and then forced a refresh of the base web site which, again, did complete in about 2-2.5 seconds... never happened again, unfortunately! Oh Well, C'est La Vie!

IF you have any questions on DNS "stuff" just ask. I've been involved in DNS stuff since the middle '80's!

I'm on the fence with respect to this new server.

Most of the time it's quite zippy. Some of the time, I see MySQL errors, something that was fairly rare before the migration (and points to a backlevel version of MySQL in their base install as "mute" noted above). I cannot explain the slowdowns you're seeing. The last of the problematic locales like VPN that I use figured it out late last week.

The grid hosting isn't working...

Overall speed: very slow, current page: 44.4s, avg: 44.9s, max: 160.1s

More hills and valleys:

Overall speed: fast, current page: 0.2s, avg: 0.2s, max: 0.6s

I think migrating to a dedicated server is the only long-term option. The idea of grid hosting sounds great on paper, but their implementation of it doesn't scale properly. I've prepared a new server and will switch the DNS entries tonight. Any suggestions for reducing the propagation delay other than tweaking the TTL value to 300 seconds? There's other TTL value like SOA TTL and secondary DNS values; I haven't fussed with them...

Dan, IF you've got some time, give me a call... I'm available from 10am until 2:30. I'll PM you with my #'s

Thanks to a pep talk from Steve, I think the DNS updates will propagate faster than last time. According to InternetSupervision, 9 of their 10 worldwide sites are getting the right server (inexplicably Washington, DC is returning "cannot resolve DNS"). The next question is whether it was worth the time, effort, and money...

Not sure if the switch has been made but I'm able to access the forum with fairly short response times at 0120 CST.

Yes, if you're reading this, you're accessing the new server (the old one only shows a "we've moved" message). If you see anything that isn't working, please let me know.

Dan, I did a couple of quick checks last night and have been hitting the site for the last 10 minutes or so, exercising forums, articles, etc and it's MUCH, MUCH better! Longest load time I've seen this morning was about 2 seconds with most of them being in the <1sec range! Mucho Improved! Thanks!

Well, I've been hitting the site here at work (where the delays were in the 20-30 second range yesterday!) and I've got to say.... THANK YOU DAN! Most of my responses are under 0.5 seconds (as verified by looking @ my proxy logs! 238ms, 307ms, 192ms and occasionally 1000ms or so, but they've been rare!). The site has been MUCH more responsive! MUCH More! I'd have to go look @ some old proxy logs to verify, but it seems that it's faster than it was BEFORE the "grid computing" upgrade...

ANY time I can help on this site, no matter how small the contribution, just ding me Dan!

Thanks, and I owe you a big "thank you"! Your tips on encouraging a faster DNS propagation worked. Less than 12 hours later and InternetSupervision reports the correct server for all their test sites. Even the VPN network I use that took over a week to update last time around is working great today.

Monday/Tuesday are the big hitter days, but from what I see so far, the response times are much more even. The unexplained e-mail notification delays are also gone.

Dan, thanks for the complement! I WOULD recommend that, in a day or two, you push the DNS TTL's back up to normal (from what you and I discussed, normal would be 1 day/86400 seconds)... No big rush, but a 1 day time-to-live is much easier on the DNS infrastructure. DNS queries are small, but there are millions of them every hour of every day...

When it comes to the web, there's no such thing as too fast. So tonight HB gets a memory upgrade just in case the upcoming Holiday Wish List contest strains the server (last year it got rather sluggish at times during the holiday). Site will be down for an hour around 1AM EST. Here's hoping the site's host doesn't screw up and crash the server!