We currently use MemSet for most of our hosting requirements. We run Ubuntu virtual servers and have very few problems caused either by us or Memset.
However, Memset don't provide out-of-hours emergency support except with fully managed servers (£40/month overhead, on a virtual server which might only be costing £10/month itself). This has occasionally caused problems in the past, with an upgrade taking place late on a Friday evening leaving the server inaccessible until after 9am Monday. It means we can't risk a major update out-of-hours in case of problems, but equally don't want to schedule downtime in office hours either.
We prefer to run multiple virtual servers rather than having one "big" one - spreading the load but more importantly meaning that one outage it less likely to hit all customers or all services at the same time (much easier to manage!). At £40/mo for multiple servers this means paying around £3-4k just to be able to contact emergency support typically once per year when we have a problem that requires it.
So, what are the current suggestions for hosts that other people can recommend with 24/7 support availability for those rare disasters? I know this type of thread crops up fairly often but recommendations change and I haven't seen the specific issue of out-of-hours support cropping up.
Other than this I have no reason to wish to move away from Memset, but we have reached the stage of migrating several sites to new virtual servers and that can either be at Memset or elsewhere, so I need to consider the alternatives.
On 9 November 2010 12:42, Mark Rogers mark@quarella.co.uk wrote:
We currently use MemSet for most of our hosting requirements. We run Ubuntu virtual servers and have very few problems caused either by us or Memset.
I'm a sysadmin/support admin at Memset - hello :)
However, Memset don't provide out-of-hours emergency support except with fully managed servers (£40/month overhead, on a virtual server which might only be costing £10/month itself). This has occasionally caused problems in the past, with an upgrade taking place late on a Friday evening leaving the server inaccessible until after 9am Monday. It means we can't risk a major update out-of-hours in case of problems, but equally don't want to schedule downtime in office hours either.
What we DO provide is that if your server stops responding to ICMP packets, this alerts an on-call engineer who will investigate for you - regardless of whether or not you have fully managed support. I know I tend to look at the support queue if this happens and the server in question is in state and I can, and do, help to resolve the issue if it's an upgrade problem and is relatively straight forward to fix.
£40 may seem like quite an overhead, but we're still cheaper than the likes of Rackspace and other providers. We manage the base OS as well (along with the standard components such as Apache, MySQL and so on) and, as you say, you get out of hours emergency SMS. We also aim to start working on your problem within 10 minutes of receiving that emergency SMS 24/7.
We prefer to run multiple virtual servers rather than having one "big" one - spreading the load but more importantly meaning that one outage it less likely to hit all customers or all services at the same time (much easier to manage!). At £40/mo for multiple servers this means paying around £3-4k just to be able to contact emergency support typically once per year when we have a problem that requires it.
Have you spoken with our sales department about this? It's not entirely out of the question that if you're looking to run multiple virtual servers that something might possibly be done about the fully managed support option. I can't guarantee it of course, but always worth having a word with them.
So, what are the current suggestions for hosts that other people can recommend with 24/7 support availability for those rare disasters? I know this type of thread crops up fairly often but recommendations change and I haven't seen the specific issue of out-of-hours support cropping up.
Not many of our competitors, that I am aware of, that would be able to do this unless they're operating a 24/7 outsourced support service and that's generally limited to rebooting only, or charging extremely large sums of money to do anything other than a re-install.
Other than this I have no reason to wish to move away from Memset, but we have reached the stage of migrating several sites to new virtual servers and that can either be at Memset or elsewhere, so I need to consider the alternatives.
Mark - have a word with Nathan or Alexandra in sales in the first instance.
Regards,
Martyn
On 9 November 2010 13:14, Martyn Drake martyn@drake.org.uk wrote:
What we DO provide is that if your server stops responding to ICMP packets, this alerts an on-call engineer who will investigate for you
- regardless of whether or not you have fully managed support. I know
I tend to look at the support queue if this happens and the server in question is in state and I can, and do, help to resolve the issue if it's an upgrade problem and is relatively straight forward to fix.
Oh, and I should also mention that we can do out of hours work by pre-arrangement. This obviously costs (, but if you do have an upgrade pending and want somebody to be on standby as a pair of hands AND don't want the fully managed support option - you just let us know and we'll be ready.
http://www.memset.com/support/self-managed.php
Regards,
Martyn
On 09/11/10 13:14, Martyn Drake wrote:
I'm a sysadmin/support admin at Memset - hello :)
I'd forgotten you were Memset! Hi!
What we DO provide is that if your server stops responding to ICMP packets, this alerts an on-call engineer who will investigate for you
- regardless of whether or not you have fully managed support. I know
I tend to look at the support queue if this happens and the server in question is in state and I can, and do, help to resolve the issue if it's an upgrade problem and is relatively straight forward to fix.
We had a specific problem a while back that I no longer recall, but the server was responding to ICMP packets but that was about all it was responding to; we couldn't SSH in, the reboot option wasn't working from the web management thing (or if it was it was rebooting to the same state), and all we could do was wait until Monday (the problem started Friday evening I think).
£40 may seem like quite an overhead, but we're still cheaper than the likes of Rackspace and other providers. We manage the base OS as well (along with the standard components such as Apache, MySQL and so on) and, as you say, you get out of hours emergency SMS.
Looking at the support matrix, there's plenty in the £40/mo that justifies it, but we don't need or want almost anything that's ticked in the managed support column. After all, "apt-get update && apt-get upgrade" handles most of the updates to the core components, and if we do it ourselves we can pick the time to do it and schedule it to suit the customers out of hours. Paying someone else to do it is fine, but we'd have to be around at the same time (to test and handle any application issues) so just arranging for a third party to run a couple of apt commands for us isn't a good use of your time or our money!
On the other hand, sometimes (very rarely) an upgrade kills something, and then we're just plain stuck! £40/mo for access to chargeable out-of-hours support is perhaps neither here or there, but if we have (say) 10 small virtual servers then nearly £4800/yr just for access to emergency support, against just £1200 for the actual hosting costs, is a bit skewed!
Have you spoken with our sales department about this? It's not entirely out of the question that if you're looking to run multiple virtual servers that something might possibly be done about the fully managed support option. I can't guarantee it of course, but always worth having a word with them.
I have, and it's kind of "sorry, this is how we do it". I have no complaint about that; Memset have no obligation to offer a service I need! But it has lead me to look at alternatives. I like Memset and want to continue using them, but I'm not able to justify paying for a range of support services I don't need just to give me access to emergency cover on the rare occasions I need it.
Mark - have a word with Nathan or Alexandra in sales in the first instance.
I've spoken to both in the past, and Alexandra today, before posting here.
I'd like to stress: I'm not unhappy with Memset's services in any way. It's just they don't quite seem to "fit" what I need. It's not a problem for me if I get a call at 3am on a Sunday morning with a problem on the web servers, but if I can't access the servers via SSH to investigate and I can't get help from anyone until Monday morning then I can't give satisfactory answers to my clients, even if that is only once every few years.
On 9 November 2010 13:44, Mark Rogers mark@quarella.co.uk wrote:
We had a specific problem a while back that I no longer recall, but the server was responding to ICMP packets but that was about all it was responding to; we couldn't SSH in, the reboot option wasn't working from the web management thing (or if it was it was rebooting to the same state), and all we could do was wait until Monday (the problem started Friday evening I think).
That to me sounds as though the miniserver ran out of RAM. Since we do not supply our VPSes with swap (as the combined shared I/O of VMs swapping out to disk thrashes the disk of the miniserver node something chronic and reduces the the performance for all). It's possible that the reboot service was functioning but, as you say, probably getting into the previous state. I assume that SSH and other services had their init scripts set to boot up on start via the update-rc.d system (or done manually)?
Looking at the support matrix, there's plenty in the £40/mo that justifies it, but we don't need or want almost anything that's ticked in the managed support column. After all, "apt-get update && apt-get upgrade" handles most of the updates to the core components, and if we do it ourselves we can pick the time to do it and schedule it to suit the customers out of hours. Paying someone else to do it is fine, but we'd have to be around at the same time (to test and handle any application issues) so just arranging for a third party to run a couple of apt commands for us isn't a good use of your time or our money!
I quite understand. I think what you'd probably want above this is some form of serial access to the miniservers so that you can remotely connect and resolve these issues yourself. We are working on making this available, but I can't give any definite timescales as to when this is going to happen. We already provide full serial console access to our dedicated servers (which also now come with DRACs which kind of negate the need for that), it's only a matter of fixing a few outstanding issues and we can provide that functionality to our miniservers too.
I have, and it's kind of "sorry, this is how we do it". I have no complaint about that; Memset have no obligation to offer a service I need! But it has lead me to look at alternatives. I like Memset and want to continue using them, but I'm not able to justify paying for a range of support services I don't need just to give me access to emergency cover on the rare occasions I need it.
I do understand, and hopefully we can resolve this issue for you in a technical capacity rather than on price or the availability of having a human being working odd hours on standby.
Regards,
Martyn
On 09 Nov 14:05, Martyn Drake wrote:
On 9 November 2010 13:44, Mark Rogers mark@quarella.co.uk wrote:
We had a specific problem a while back that I no longer recall, but the server was responding to ICMP packets but that was about all it was responding to; we couldn't SSH in, the reboot option wasn't working from the web management thing (or if it was it was rebooting to the same state), and all we could do was wait until Monday (the problem started Friday evening I think).
That to me sounds as though the miniserver ran out of RAM. Since we do not supply our VPSes with swap (as the combined shared I/O of VMs swapping out to disk thrashes the disk of the miniserver node something chronic and reduces the the performance for all). It's possible that the reboot service was functioning but, as you say, probably getting into the previous state. I assume that SSH and other services had their init scripts set to boot up on start via the update-rc.d system (or done manually)?
Hmmm, all this sounds to me like there's no equivalent to console access on your VPS systems, which should have then given them more to play with. With Mythic's Xen offerings (used to be bluelinux's), xen-shell access is provided so that you can see the console of the vps in question, and fix most any problem and get full diagnostics. Maybe this is something you could consider in your offerings?
Thanks,
On 9 November 2010 14:21, Brett Parker iDunno@sommitrealweird.co.uk wrote:
Hmmm, all this sounds to me like there's no equivalent to console access on your VPS systems, which should have then given them more to play with. With Mythic's Xen offerings (used to be bluelinux's), xen-shell access is provided so that you can see the console of the vps in question, and fix most any problem and get full diagnostics. Maybe this is something you could consider in your offerings?
Indeed - we're working on offering Xen console access to our customers in the near future.
Regards,
Martyn
On 09/11/10 14:05, Martyn Drake wrote:
I assume that SSH and other services had their init scripts set to boot up on start via the update-rc.d system (or done manually)?
Yes. We have had memory issues more recently but I don't recall this being memory; I think it was something broke after an update, but I could be wrong (that's definitely happened on other occasions).
The fact is that we've only had probably three serious issues like this in several years with Memset, but the problem is that it makes it difficult to schedule routine updates out of hours just in case something goes wrong. Doing them in work hours isn't acceptable to customers, so the temptation is not to perform updates at all (actually we tend to do them 8am Monday morning so we're not too far from support being available but it's not an ideal time for lots of reasons at our end).
I quite understand. I think what you'd probably want above this is some form of serial access to the miniservers so that you can remotely connect and resolve these issues yourself. [...] it's only a matter of fixing a few outstanding issues and we can provide that functionality to our miniservers too
Yes, that would be helpful. I don't suppose you have any idea of timescales (I won't hold you to them!)?
I'm also told by sales that some kind of out-of-hours for self-managed customers is in the pipeline, but that is probably months away.
[I assume that this isn't of interest of ALUG so unless someone says otherwise further replies from me will be off-list.]
On 9 November 2010 14:23, Mark Rogers mark@quarella.co.uk wrote:
Yes, that would be helpful. I don't suppose you have any idea of timescales (I won't hold you to them!)?
At this point, I don't know. But we are looking to get this out as soon as possible. Weeks rather than months I would hazard I guess.
I'm also told by sales that some kind of out-of-hours for self-managed customers is in the pipeline, but that is probably months away.
I believe this is the case, but again, I doubt this is going to be that far off.
Regards,
Martyn
-
Brett Parker -
Mark Rogers -
Martyn Drake