New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
@Francisco is a great guy with the rare blend of hard skills necessary to run a hosting provider with the soft skills to engage customers in a mature, responsible manner.
Every time there is an outage, you can expect a detailed RFO from @Francisco if you ask for it.
But if you look at Buy* threads on LET, it's the same scenario every time: complaints about rude/incompetent support, outages, etc, and then @Francisco comes back with pure transparency, competence and 110% customer satisfaction. And this happens thread after thread after thread.
With Buy* you get more than what you pay for, but you don't pay much to start. So you don't get Mailchannels for reliable email delivery, you don't get stability, but you always get premium @Francisco no matter what.
Does that domain info have to be in httpd.conf? I remember there was a trick with the apache rewrite module to implement vhosts by getting the domain name from the client's host header and looking it up in a mysql table. So you'd load all your domain names into a local mysql and keep it running, and then restart httpd as often as you want without having to reload the domains. Maybe litespeed can do something similar. "apache mass virtual hosting" is a good search term.
Maybe there's a plugin for it but it wouldn't work for Litespeed nor would cPanel have it integrated.
Anyway, a few users already agreed to lower their count to < 1000 domains per node, meaning one node already fell by nearly 8,000 alone. I'm keeping a close eye to see if we have anymore quirks with this. Faster single thread CPU's would've helped (assuming that's what this issue is) but it would've just been masking it, it wouldn't be 'fixing' a bigger issue.
At some point all those sites want SSL renewals though and they're going to require a graceful restart for each of them. LetsEncrypt-for-cPanel has a 'deferral' option but even still.
@MikePT said he has forwarded a 'cap domains a reseller can have, total' feature request to cPanel since there's no way to even track this information. I have a script that helps build that data, but not everyone is that lucky.
Francisco
What, you can't dynamically swap the certificates around? You should complain to litespeed devs. Your use case has to be important to them so they should fix it. Also I'd ask them for a better solution to mass vhosts in general. In your environment it's not nice to have to keep restarting.
You're mistaken, I wasn't asking you to, I was making a point.
That ^^ is being a dick.
It has to reparse the entire Apache config. Litespeed does its restarts gracefully and usually it does them w/o any sort of blip. Sometimes it can't finish (or it dies) and it has to start from the beginning.
In 5.0/5.1 they did a major overhaul of their parser which did speed it up a lot. Still, making it run through a 70MB configuration file sounds like it could be a serious problem since there's a lot of stuff that cascades and the like. As for the SSL, cPanel doesn't integrate via Litespeeds own vhost manager, Litespeed just reads the Apache configuration and goes from there. The only way it'll know about SSL changes is if it sets up file watches or just waits for graceful restarts.
This is all still a 'maybe'. We'll see how many blips we get in say, the coming week, and see if it's still a problem. If it's fixed? Then hopefully cPanel merges it as a fix for these types of abuses. You're all welcome for that feature.
If it isn't helping then it's still good to get that feature in, but there's still something bigger causing issues (at least for our size deployments).
Francisco
@Francisco have you even patched + rebooted your nodes yet? My VPS shows 30 days uptime making me wonder if I am vulnerable to Meltdown + Spectre. . .
I'm not sure but he mentioned in a comment regarding it. He'll wait a week (from that time) and see if there's any major issue comes up due to these fixes before starting to patch nodes
Well that patched OpenVZ kernel was released a week ago today. . . (though I am on a KVM node, for which the kernel patch was released 1 week + 1 day ago)
No ETA for this, but after speaking to you, it does make sense to implement such limits, reseller wide that'd apply to subaccounts as well.
It's non sense to have, say, a 20 USD reseller hosting 10k domains, that'll definitely slow the server down, and reseller acls do not support such limiting for now.
I did report that to a cPanel dev but I'm still waiting to hear from him.
I've been waiting to see if there's any big problems with the patches as well as waiting for the BIOS updates.
Francisco
The existence of a 70MB config file tells me something is broken about that setup. Maybe the brokenness is forced by how litespeed or cpanel work but they should really be thinking about a fix. Configs are something you set up once and mostly leave alone. That 70MB of stuff should be in a database, not a config.
Well, cPanel didn't plan for Amazon affiliate sites loading up 20,000 sites, but maybe they will if they hear from a few more people that it has been a problem.
We had a case a couple years back when we first started where a user had literally 32,000 sites throughout all of his $7/month reseller. He had InnoDB databases for each so in the end the user was accounting for something like 60GB of innoDB buffers.
As I said, we don't know if this is the source of the blips, but it isn't helping.
Francisco
Yikes, lol. Hope that's fixed, especially the innodb.
Well, let's see if the recent changes help. Otherwise I don't mind tossing out more troubleshooting suggestions if you think they are useful.
The user was told they were abusing resources. They argued otherwise so we simply packed their files and refunded the months service.
Francisco
Stop getting-shit-on-blocking.
#muddydicks
I am using Frantech services (VPS and shared) since 15 years and it's my best hosting ever (of course only better is http://mzunguhosting.ml/). Cheers!
I can't understand why we can't just have a node for non reseller accounts with a 100 domain limit. For the $5 a year price 1gb and $10 5gb it's more than reasonable. I would imagine you would prefer a ton of happy $5 a year customers with less than 100 domains than a few happy $20 a year reseller customers with 1000+ domains constantly bringing down your network and requiring more and more of your personal time which is undeniably very valuable time.
Your customers are literally telling you they would prefer more stable service option with a domain limit of a seperate node from the abusers and your saying "unlimited" domains sells better... Well down time doesn't sell anything and a 4 page post of people frustrated and concerned should tell you that.
I see your point but sounds like that what would happen is that this node would end up having more clients, as opposed to same with less domains per client. Same end result.
Unless @Francisco couple other limits to the node as well, besides the X domains per account.
Unfortunately I have to agree about the power of "unlimited" marketing. I have had endless arguments about how it's better to pay for what you actually use, since 1) there's no such thing as unlimited resources, 2) most unlimited plan users use less than the average and subsidize a few accounts who use a lot, or 3) "unlimited" is never truly unlimited and is limited by fair use policies, etc.
Even when people agree with these points, they STILL can't shake their preference for "unlimited" offers. Behavioral economics/psychology ftw.
Everything is relative including this. At a price point of $5-10 a year you're dealing with two primary groups. Those that want hosting for simple sites and need a basic webhost and those that don't care if they get kicked off and intend on abusing it.
So with this in mind if your part of the first group that just has basic sites that need a web presence then 100 domain limit will easily match what you need for the price. However if your part of the second group and you want to setup 1000+ amazon affiliate link sites with intention of using it into the ground until your kicked off then the 100 domain limit would be a stopping point.
Besides I'm not asking for him to remove the unlimited plan just give the option for sign ups. 100 domain limit on a separate server with like minded users or unlimited with like minded users. I know @Fransico has enough left over gear that he can easily put together another server.
Well when he's selling a $20 a year reseller account and ends up hosting 8000+ domains on a single account I assume he would prefer to have 80 $5 a year (100 domain limit) accounts hosting 8000 domains instead ($400 a year, literally 20x more revenue). More revenue equals greater ability to build out the nodes keeping over utilization low.
Resellers and shared users don't mingle, they do have separate nodes.
Unless you have a really old account and somehow got missed when we did our big reorganizing years back, you will be in a designated place.
Francisco
On one node alone we have a half dozen resellers accounting for ~75% of all domains on the node, causing our httpd.conf to be ~70MB in size instead of, say, 5MB or so where it should be. This obviously will cause any webserver to start up slow.
We can't seem to track down that common down time experienced on the non-reseller node...
Then perhaps the shared users also need to have limits opposed. I don't think I need to post the downtime records for December to prove a point but I will if it's needed. We rag on providers in this sub like snipped and others for overselling into unreliability. At some point buyshared.net would also need to be held accountable.
Have you mentioned anywhere which node you're on? If you have a shared account you should be on either lv-shared04 or lu-shared04. If you have a really old account (from our starter year) you might be a left-over on lv-shared02 before we split the node up.
Again, you're assuming that this is related to the domains. The amount of domains on lv-shared04 shouldn't be an issue.
Francisco
I'm on lv-shared04. If this isn't an issue with overuse of accounts on a shared platform then we have bigger issues at hand here. Because frankly then it's more to do with the management and setup of the service than over-provisioning. I had a site down for 5 hours the other day, then a nearly daily downtime of 20 mins. If this isn't a result of large conf files taking 20 mins to restore than what is the issue?
You're paying anywhere from $5/year to $15 for individual, and $2/month to $7 for reseller. The "stable" option is to pay more elsewhere because obviously @Francisco is approaching this market from the low end and packing these nodes up.
Will you accept a refund so this thread can be closed? Its a budget service and for $5 year bashing the provider since days its not fair. Just accept a refund and close the thread please.
Those were likely due to the ddos-protection features we had enabled on our juniper. It killed off a bunch of IP's on our shared nodes. We fixed that the first time but then it happened the 2nd time (the 10 - 15 minute one you saw) at which point we simply disabled the feature all together.
Not all IP's were out, each time was around 20% of all IP's on the node lost their ARP entries.
I'd have to go through my logs but I don't think l*-shared04 has the blips that the resellers do, at least not as often. I'm pretty sure it does it once in a while but I can usually find a log of wswatch forcefully restarting it since it crashed.
Francisco