Stalled web pages, MySQL errors and more

Customer Support Customer Only Discussion
1

Since the move to Waco, I have noticed an improved responsiveness on my clients’ websites. That is good as the websites no longer seem slower than sites hosted in NZ.

However, since the move, I have noticed quite a few issues which seem to be ongoing and give the impression that the hosting service I provide to my clients is unreliable.

Websites frequently slow to a crawl or simply stall, frequent “Connection closed by remote server” messages, MySQL error messages, unable to log into control panel etc. Most of these problems are very short term, and by the time the client reports the issue to me it disappears. However they happen often and my clients notice. Some have suggested that reliability has been compromised for speed. They would prefer to go back to “slow and steady”.

Originally I was able to get away with the explanation that it was teething problems after the move of data centres, but that is wearing a bit thin after this time. Several clients are making definite threats to move due to the erratic behaviour of their websites.

As I mentioned, most problems are short-lived, and usually disappear before I can report the problem. I have felt it pointless to create support tickets as when I have, the response is generally “it is fine now. Please check again”. Is there any point in creating tickets in these circumstances?

Is there some reason why these problems are occurring now at Waco when they didn’t occur in Florida? I had been extremely happy with the service provided by Jodohost until the change of data centres. While on the whole, support is still good, the hosting service isn’t. I’m starting to wonder if it’s in my best interest to remain faithful to JH.

2

Can you please PM me or Pratik the tickets where your clients are seeing frequent problems?

3

I am not sure about the unable to login to control panel, except right now, where there is some odd issuing starting up hsphere, and it occurred while Parallels was in working on another issue (unrelated to license, about payment gateway support)

Linux, we’ve got some fixes rolling out soon, we are aware of some issues there and checked, and have resolution, it is not a ‘simple’ drop in fix, but a rather comprehensive one, and will be rolled out soon.

I am unaware of any long term ongoing issues on the windows side, we’ve got a few here and there, posted, and addressed, MSSQL8 has been a bit of a pain of late I admit.
However, overall I am not mostly seeing more issues than before, as before problems were starting to increase with hardware issues, which while still a few hardware issues here and there now, it is not as bad as it had gotten before.

I think once we get the linux web server fixes rolled out, we’ll be seeing a big positive change on the whole. And yes, do let Pratik, Tanmaya, and myself know of any issue tickets so they can be addressed.

4

A quick reply as at the moment, these forums. resellerdesk.jodohost.com on a definite go slow and the CP is not working. Websites on multiple servers - web5, web10, web11 and possibly others (not sure as can’t access CP) - are taking minutes to download or timeout, yet other US based sites are at normal speed. I’m wondering if there is some routing issue between NZ and Waco which didn’t exist between NZ anf Florida?? Will provide more info when things working better.

5

there forums, are at softlayer, not in Waco.

Seems routing maybe?

and CP is up for a bit now :smiley: Thankfully!

6

Actually there were routing problems last night NZ time. Shortly after my previous post here, all overseas sites ground to a complete halt, not just those at JH. I guess the problem had been getting worse during the day, we were unlucky in that JH sites seemed to be affected most. It was my frustration last night that prompted me to start this thread.

Unfortunately it is very hard to quantify the degree of the problems I have mentioned. The Internet, being what it is, one can expect web pages to fail to download correctly from time to time - especially here in NZ being at the end of the internet world so to speak. Prior to the move to Waco, there didn’t seem to be any difference between JH hosted sites and other internationally hosted sites. Since the move to Waco, I, my customers, and our website visitors notice that one has to hit the browser refresh button more often with Jodo hosted sites than with other sites. They also notice the erratic download times for web pages. While in general pages appear to load faster, we all notice that there are more frequent occasions when a page loads very slowly - taking over a minute for a page that normally loads in a few seconds. Again, this does happen with any site from time to time, but the perception is it occurs more often with JH sites.

I want to emphasise that these are still relatively infrequent occurrences, but my customers tell me that they occur more often than before the move, and more often relative to sites hosted elsewhere. Because this is a general perception, rather than specific incidents, I mostly haven’t bothered with trouble tickets. It’s very hard to be specific, which is one reason why I posted here.

I can provide only a certain amount of spin to my customers before it ceases to be effective. I’m not sure what, if anything, can be done to remedy the situation - it’s probably outside of Jodohost’s control. I am hoping that someone can say that there is something that is being worked on that will improve the performance of Jodo hosted sites. But I’m not holding my breath…

7

Please read my first post in reply, we are rolling out some changes on the linux side that should help. We’ve got web7 on cluster2 as the test bed and it has been rock solid since finalizing the deployment, and about 48 hours after with some tweaks. I don’t knwo of any ongoing issues on windows, there are the occasional app pool recycle needs, and we can move app pools if needed, otherwise the linux fix rollout will probably help in these cases you mention. It requires a custom compiled Apache, PHP, CGI stack on each server, so it can’t just be done overnight.

Last night sounds like a abnormal event and should probably be throw out of the equation, but as said these forums are actually off network, and it saw the issues too, so it wasn’t us.
We actually have better routing to Australia now than before, if may help me if you can get winMTR on your machine and do some traces with winMTR when things are ‘normal’ and then when things don’t work right as well, then we can see if there is some increased packetloss, or latency.

Early after the move one of our locals down under (antic), noticed an abnormal event with some routes that hit he.net in California, we were able to isolate this and finally get a fix in place, it did pop up about 2 months later but we reported and got resolved again. This is out of our direct control in some manners, but if ongoing for a long time we can route around them by basically not allowing traffic to use he.net, since there are many other providers in the mix.

Add to that, we are now getting multiple 10gig core links back to Atlanta and Miami to mesh in with the FPL Fibernet network that we had before in Miami, and we’ll have honestly, the best of both worlds, able to route around the world east, west, north and south all from here.

Right now our facilities are linked to Dallas, Austin, San Marcos, Houston, (all Texas, there are also some other Texas cities with non core markets not mentioned), and Washington DC. Soon Atlanta and Miami will be added in the direct connection mix. FPL was installing the fiber and switches for this last week.

8

Steven, you have provided me with the information I needed to hear. All my clients are hosted on Linux servers, so it’s good to know that work is under way to improve performance.

You’ve also mentioned improvement to external links and I’m hoping these will see some improvements too. I seem to recall some discussion in the forums about the problem antic experienced and am glad to know that you can even affect some changes not under your direct control. I will keep this in mind for the future.

Again I want to point out that if it wasn’t for the frustration I was experiencing yesterday evening, I wouldn’t have started this thread. At that time it appeared to be Jodo related rather than all international connections. I apologise if this came across as a complaint instead of an enquiry. Query has been answered :slight_smile:

I have already thrown last night’s problem out of the equation as far as Jodo is concerned.