Server Down?

Discussion in 'H-Sphere Shared Hosting' started by TekniDude, Nov 15, 2003.

  1. Hawkeye

    Hawkeye Perch

    I'm quite skeptical of these figures...especially considering below;




    I'll be happy to comment - we get about 1000 unique visitors per day, which can rise to about 1400 on the most busy days. In the world of web hosting this is really considered a mid-sized site at best and not a world beater. We offer no streaming audio, video, or software downloads. Our "Wallpaper" section held our largest graphics, which totaled about 40 and ran about 800K each. Of course, those were not constantly downloaded and were not the most popular area on the site (the WWG forum was).

    We spent the past 6 months with a "Premium" hosting service, who NEVER made issue with any of this. I simply switched to Jodohost for a better deal (cost wise), but now I see that was a big mistake.

    Also, as a point of clarification, we do not use SQL for our site, but run on an Access database, which is a WHOPPING 3.5 (that's 3 and a half) megs in size. Wow. Sure, Access may be CPU intensive at times, but a database of this size?

    What has really happened here is that ANOTHER customer installed PHPBB on the server this past Saturday, which brought it down 12 times in a 24 Hour period. Support identified that person and and disabled their account. After that, they started scanning their logs for other "high utilization" customers and shut them down also, which is what happened to me. Bad, bad business practice.
  2. Yash

    Yash Bass

    You ARE generating high usage. I have system usage reports when your site was running and when it was taken offline.

    I believe it is your access database then. Most of your visitors may be visiting during a certain period of time resulting in a large number of queries being generated at that time. We had been monitoring your site. You WERE bringing up CPU usage to 40% + for a sustained period of at least 10 minutes
  3. FreeWilly

    FreeWilly Perch

    hmmm.
    I do hope JodoHost revises this termination policy. I'd like to be provided some notice before termination

    I do however understand JodoHost's stand on server usage. The speeds and uptime around here have been excellent and Yash did commit they they would maintain that for years to come.

    Hawkeye, MS Access is very inefficient if you are getting 1500 visitors a day. If a good number of them visit during a 10 minute period (like Yash says), it can cause really high CPU usage from my experience with MS Acesss.
  4. Ron

    Ron Perch

    I've not commented on this but I will now.
    I am disappointed that JodoHost has been this brutal with high usage customers. Some notice is standard business practice. I've really never seen such a controversy here at JodoHost for a long time. They've always been very professional with PR.

    I get 1800 visitors a day. My site runs MS SQL. Thankfully my site hasn't been disabled. I think that is why MS SQL is far more recommended as back-end for high usage websites. I totally agree with FreeWilly that MS Access indeed can cause high usage.
  5. brawney

    brawney Perch

    SQL Server vs Access. This sheds some light on things. I'm surprised that the site was even working at all with that many people on an Access database. We tried using Access for a small web site and it didn't handle multiple users very well at all. And I'm talking about a couple users, not hundreds.

    So it sounds as if we're using SQL Server on our portal that we will be OK. That plus SQL Server being on a different box in the new cluster sounds like we'll be in good shape.

    My site will never have that much traffic so that eases my worries. I have also been very impressed with JodoHost's performance since joining a couple months ago.
  6. eko

    eko Perch

    How bizar, first MS SQL now MS Access what next. The principle is that interactive, dynamic database driven sites are taken down, not because of bandwith- or diskspace usage but processor usage. Although Yash mentioned the occassional practice by other web-hosts to take down sites he forgets to mention how this affected that web-hosts business.

    It's also strange that seemingly popular sites are taken down as many hosts would actually make deals with those people, including advertising.

    Most incomprehensible is that Asp Invision (hosted by gearhost) advertised jodohost with a banner on its main portal site. Now that the asp invision portal seems to be part of the problem I think that jodohost should take part of the blame as it obviously gives the impression by advertising on the asp invision portal's site that it supports the portal software to be installed on its servers and used.
  7. Yash

    Yash Bass

    We have SOOO many customers using MadASP portal. This is only one customer. I believe the MadASP portal is finely written code. But if you are going to have a really popular website and run MS Access as backend, high resource usage is inevitable

    MS SQL also causes high CPU usage. But that normally happens on large databases when you are banging the server with complex SQL queries every other second. Yes, we have had a customer that was doing this.
  8. Yash

    Yash Bass

    We have plenty of interactive, dynamic database driven sites. Only a few generate such high processor usage. In those cases we have no choice to but disable it in order to protect other customers. Web hosts frequently disable customer sites that utilise high resources. It is only because we have a forum here that is gaining so much attention
  9. brawney

    brawney Perch

    You make a good point Eko. It sounds like JodoHost was almost guessing at why Hawkeye's site was causing high utilization. I would think they would at least know for sure whether it was SQL Server activity or Web server activity or both. Makes you wonder about their monitoring techniques.
  10. Yash

    Yash Bass

    There is no monitoring technique available. It depends on the skill of the administrator. It is so hard to find websites that are responsible for crashing or resource problems. But when we isolate a website, we monitor it. I've repeatedly stressed even in the case of daddyc is that we don't know exactly what is causing the high usage. But we did conclusively determine it was because of the customer's website.

    Hawkeye said he was using a database, I assumed it was a MS SQL database. An MS Access database explains the situation nicely though
  11. brawney

    brawney Perch

    Thanks for the clarification Yash. Access really isn't a good choice for a web site with any real amount of traffic. Perhaps with SQL Server being on a different box in the new cluster it will not only help performance issues but will also separate the logging so you can have more information about why and where a site is using high levels of resources.

    Thanks again.
  12. eko

    eko Perch

    Hi Yash,

    Thanks for acknowledging that your monitoring systems are not precise and can't pinpoint the problems beyond the problem causing "domain/customer". This should make you more cautious and relates to my earlier statement that:

    "In legal terms it seems that your customers can be found guilty, charged and sentenced without any warning, defence or probation."

    This victimisation of customers is what worries me most, especially as you claim superior service. Some postive action on your site will make me feel less unsettled.
  13. Hawkeye

    Hawkeye Perch

    So, it still sounds like a customer with a busy site, on Access or SQL is going to get booted.


    Agreed. If it was determined that my site was causing the issue, and in particular, the use of Access was causing the issue, I should have been moved to a different server for some limited time and at least been given the opportunity to upgrade to SQL, or take other action. This would have resulted in customer goodwill and no outage for myself, or any other users. I was never given any warning or indication that my site was causing high CPU utilization and there is no way for me to monitor this.

    Instead, I was sent a letter AFTER my site had already been taken down and I'm now 2 days into an outage which will last at least a week or two while I try to find another host.

    At this point I know of two users who just had their accounts cancelled via this thread, in the past 2 days. Were there others? The point is that this could have all been handled much better.
  14. Yash

    Yash Bass

    1) A busy site that generates such high usage will be kicked off. any host will do this

    2) We are not victimising the customer. The customer's website was victimising the servers. We brought down the site temporarily. After we determined conclusively it was causing usage problems, we notified the customer. Once the customer replaced his site with an HTML page, we reactivated it. We also will refund 2 month's of service to this customer once he finds a new host. Our TOS doesn't require us to give any refund. I'm willing to offer all the assistance I can to help this customer shift to a new host.
  15. eko

    eko Perch

    Yash, the problem is with your answer. You don't allow high usage sites, fair enough when you mean high usage translated into processor usage not bandwith nor diskspace (for which you can charge extra in any case). Your customer however indicated willingness to change the possible causing MS Access database to an MS SQL database, but your reply is to change host. That's victimisation, there's no consultation only one-sided action, and that's where the worry lies. You guys do a great job, but customer dialogue is essential for any good business.

    Finally in an earlier message you wrote that you support plenty of dynamic interactive sites. Could you please provide us with examples of those sites. It might give us an idea where the ceiling is. I don't think a site with 1,400 users is big, but you seem to.

    It's a bit worrying that you monitor your sites using "oh, that's nothing. We use the most advanced technology ever to monitor our servers. It is called the "ping" ;)" and "yep, nothing beats that 16-bit DOS ping program" as posted on 6 October 2003 at http://forums.webhosting.info/printthread.php?t=177&pp=40

    This, given the current debate, doesn't give me much confidence in that you actually manage to monitor your servers that well.
  16. Yash

    Yash Bass

    Point noted.

    hawkeye and teknidude, you are both free to re-examine your site. If you can determine what was causing your sites to generate such high usage, please email [email protected]. You are free to correct the problem and we'll turn on your sites and see how they behave. If they don't generate further usage problems, you can continue services with us.

    I'll have an official email sent out to all customers with disabled sites.
    But the policy of disabling sites first is just so necessary. The more we allow such sites to run on the system, the more we risk a crash or other problems.

    Both customer site's are active BTW.
  17. Yash

    Yash Bass

    There is no such upper limit. A site getting 10,000 visitors a day may not generate as much processor usage as a site getting 2000 visitors a day. it depends how the site has been designed.
    We have a strict privacy policy so I can't let you know of any sites we host. I can say however that we host alot of portals, commercial sites and even a web hosting directory. This forum has plenty of urls mentioned

    eko, surely you don't take that post seriously :)
    Currently, our datacenter monitors our servers for us. We are immediately notified on downtime.
  18. eko

    eko Perch

    Now that's what I call positive action! Thanks for listening. I'm curious how those problems get resolved so that we can avoid making the same mistakes. It might be a good idea to make a "record" of possible known difficulties (for example: MS Access database calls taking too much processor time when using ASP invision portal with more than x-amount of users, you're advised to use MS SQL or contact jodohost support)
  19. Yash

    Yash Bass

    Yes, we'll look at more effective ways to warn customers about high usage problems before it comes to a point where we have to disable that site. This is all planned for the new cluster we'll have up shortly.

    Win5 is going to be pulled down after all sites are transferred off it.
  20. Hawkeye

    Hawkeye Perch

    Based on what the email from customer service told us to do, I cleaned out (deleted) all of my files from the server yesterday, with the exception of a single placeholder default file which just tells our visitors that we are down and working on a resolution (support did reactivate that file for me). Since that time I have been researching new hosting options.

    I can try to upsize this database to SQL, or run in some limited fashion until I am able to successfully do that, but eventually I expect to be getting the same number of daily hits again and I'm worried that all this could happen again, even on SQL, based on earlier comments (you have also shut down at least one SQL site before, correct?).

    So, after what has transpired, I am a bit apprehensive. I also need to understand what is offered in the way of customers backing up their own SQL databases. I guess I would like to find out, based on what we know about my site's usage, would it work here, or not, even as an SQL site? (Are there other SQL sites of this size or larger running successfully at Jodohost?). If not, I need to make 1 move now and convert to SQL with the new host (verses going SQL here, and still being booted again a month from now and having to move anyway).

    Any insight would be helpful...

Share This Page

JodoHost - 26,000 hosting end-users in 100 countries
Plesk Web Hosting
VPS Hosting
H-Sphere Web Hosting
Other Services