What do you require from a company that does manage a server for you (let's say it is a dedicated server)?You mean in the way of what services they should provide to manage your server? Not following.I mean for example:- What should they manage?- What service should they provide?- What should they do on request?- How fast should they start with fixing a problem when (for example) apache crashes or there is a hardware failure?- How should they know there is an issue?So: what do you expect from a provider that offers a fully managed dedicated server?I'll expect them to do all server admin work that I would do if I need to manae the server. OS Install and Update, firewall, antivirus, antispam, managing e-mail server, Adding some software applications I need to use, and etc.Server management companies should alway offer good security including rootkit, firewalls, brute force detection, php modifications and many other things that will help secure your server. They should also offer 24x7 server monitoring with auto reboot. Also look for guarantees on ticket response and resolution times. Answering tickets in a timely manner is important in the server management industry.Why do you consider "auto reboot" a good thing? Why should your server auto reboot? Why is it going down in the first place?That's something that still puzzels me to this day, why is "auto-reboot" good? All it says to me is that your server goes down often and that in my opinion is a problem, the day "auto-reboot" is an option is a very bad day.Maybe I am missing the point?Auto reboot is a great option for a management company to offer. If the server for some reason goes down (doesn't mean it happens a lot) then it can be rebooted immediately. This option can sometimes offer a faster reaction time on getting the server back up and running.It still misses the point of why did it go down? Things don't just fail to the point that they need rebooted.I'm going to have to agree with Scott on this one. Perhaps you're confusing "server monitoring with auto reboot" with "server monitoring with response".
Just because I monitor a server (or service) as "down" doesn't necessarily mean the entire server is down.
This is one of the major flaws with SIM in fact. Creating an "auto reboot", or "auto restart" program like this just makes it that much easier for you to be DOS'ed. Think about it a second:
attacker finds a way to flood apache with requests, causing SIM to think apache is down, restarting apache.
This done enough times will cause the server load to go up, and thusly trigger the "auto reboot" action, causing more downtime. All the while, your attacker is laughing his (or her) tail end off, because you were dumb enough to put such ridiculous software into play in the first place.
Now, don't get me wrong, I'm not saying that management don't need to address issues, but auto ANYTHING is bad. That simply gives a server the power to kill itself, or assumes (incorrectly) that the server is smart enough to know what's up. It's not!Scott.Mc, As you probably know sometimes things happen like the server locks up because of a clients broken script(or something like that) and having the peace of mind that the server can be rebooted when needed helps many people sleep with out worry. I can see where your argument is though. A management company should be there to catch the issue before it happens any way.The real answer to your question is going to be based on a number of factors including the amount of revenue you're trying to protect, the level of hands-on support you require, the number of incidents in a given month, the complexity of your system.A lot of server management companies are essentially a ticket queue of systems administrators with a couple of years experience. They seem to be find for simple issues like software installation, service restarts, server reboots, ticket requests. Most of these companies work within the constraint that you must use a control panel, often specific control panels. This makes a lot of sense, because it's the only way they can easily enforce standards amongst a wide customer base. They know that whatever control panel du jour you're using is going to write configuration files and behave the same amongst every customer. Essentially what you're saying is "I don't need a full time junior systems administrator for my handful of servers. I just need a junior systems administrator to spend an hour or less a month for my minimal needs" and it's a great model.However, as your application or needs increase in complexity, or in risk, then the ability for these companies to manage your needs approaches zero. This model breaks down as your needs increase in several arenas, including: - When you're in need of in-depth architecture guidance- Performance and Scaling issues requiring understanding between the interactions of disparate applications- Maintenance of high levels of reliability while increasing load characteristics and complexity- Increased regulatory requirements such as PCI and CISPIn the end, you get what you pay for. In highly competivie metro areas like San Francisco or New York, Senior Systems Administrators range between $110k and $160k. Even the Senior Admins I know in secondary and Tertiary markets like Austin and Madison make $90k-$120k.Scott.Mc, As you probably know sometimes things happen like the server locks up because of a clients broken script(or something like that) and having the peace of mind that the server can be rebooted when needed helps many people sleep with out worry. I can see where your argument is though. A management company should be there to catch the issue before it happens any way.What if the issue cannot be predicted, and is only disco, such as a linux kernel panic? Then the appropriate procedure is :1. Go back in time, make sure you have a Netdump server, and have the netdump client correctly configured on your server, as well as MagicSysRq2. While back in time, configure your server to automatically dump core to your netdump server, and also to reboot after the dump is successfully transferred to the netdump server3. Make sure you have some sort of trigger system on your Netdump server that generates a ticket when a new dump is in your tftp directory4. Begin root cause analysis with the kernel dump with gdb and other related tools. Using this as part of my common procedures, I've worked at insane, very high volume startups like Napster where I was the only systems administrator with 500 mostly identically configured servers under my watch. Automated systems administration is the only way to really scale.The controversial bit here is obviously part of step #3. You might want to require manual intervention, which would men you still need MagicSysRq configured, as well as a server with a LOM or remote serial access. Then you'd need to console into the server, force a manual netdump, and reboot the server.Just "Full Managed". So they will keep your server up2date. And manage your server, install things for you etc. That are my requirements if I will choose for a Managed Dedicated.But I don't have managed dedicated for my self
Its a good option to select a perfect package for your server and then signup.. that way you know what you are going to get and what you want to be done on your server.
Just because I monitor a server (or service) as "down" doesn't necessarily mean the entire server is down.
This is one of the major flaws with SIM in fact. Creating an "auto reboot", or "auto restart" program like this just makes it that much easier for you to be DOS'ed. Think about it a second:
attacker finds a way to flood apache with requests, causing SIM to think apache is down, restarting apache.
This done enough times will cause the server load to go up, and thusly trigger the "auto reboot" action, causing more downtime. All the while, your attacker is laughing his (or her) tail end off, because you were dumb enough to put such ridiculous software into play in the first place.
Now, don't get me wrong, I'm not saying that management don't need to address issues, but auto ANYTHING is bad. That simply gives a server the power to kill itself, or assumes (incorrectly) that the server is smart enough to know what's up. It's not!Scott.Mc, As you probably know sometimes things happen like the server locks up because of a clients broken script(or something like that) and having the peace of mind that the server can be rebooted when needed helps many people sleep with out worry. I can see where your argument is though. A management company should be there to catch the issue before it happens any way.The real answer to your question is going to be based on a number of factors including the amount of revenue you're trying to protect, the level of hands-on support you require, the number of incidents in a given month, the complexity of your system.A lot of server management companies are essentially a ticket queue of systems administrators with a couple of years experience. They seem to be find for simple issues like software installation, service restarts, server reboots, ticket requests. Most of these companies work within the constraint that you must use a control panel, often specific control panels. This makes a lot of sense, because it's the only way they can easily enforce standards amongst a wide customer base. They know that whatever control panel du jour you're using is going to write configuration files and behave the same amongst every customer. Essentially what you're saying is "I don't need a full time junior systems administrator for my handful of servers. I just need a junior systems administrator to spend an hour or less a month for my minimal needs" and it's a great model.However, as your application or needs increase in complexity, or in risk, then the ability for these companies to manage your needs approaches zero. This model breaks down as your needs increase in several arenas, including: - When you're in need of in-depth architecture guidance- Performance and Scaling issues requiring understanding between the interactions of disparate applications- Maintenance of high levels of reliability while increasing load characteristics and complexity- Increased regulatory requirements such as PCI and CISPIn the end, you get what you pay for. In highly competivie metro areas like San Francisco or New York, Senior Systems Administrators range between $110k and $160k. Even the Senior Admins I know in secondary and Tertiary markets like Austin and Madison make $90k-$120k.Scott.Mc, As you probably know sometimes things happen like the server locks up because of a clients broken script(or something like that) and having the peace of mind that the server can be rebooted when needed helps many people sleep with out worry. I can see where your argument is though. A management company should be there to catch the issue before it happens any way.What if the issue cannot be predicted, and is only disco, such as a linux kernel panic? Then the appropriate procedure is :1. Go back in time, make sure you have a Netdump server, and have the netdump client correctly configured on your server, as well as MagicSysRq2. While back in time, configure your server to automatically dump core to your netdump server, and also to reboot after the dump is successfully transferred to the netdump server3. Make sure you have some sort of trigger system on your Netdump server that generates a ticket when a new dump is in your tftp directory4. Begin root cause analysis with the kernel dump with gdb and other related tools. Using this as part of my common procedures, I've worked at insane, very high volume startups like Napster where I was the only systems administrator with 500 mostly identically configured servers under my watch. Automated systems administration is the only way to really scale.The controversial bit here is obviously part of step #3. You might want to require manual intervention, which would men you still need MagicSysRq configured, as well as a server with a LOM or remote serial access. Then you'd need to console into the server, force a manual netdump, and reboot the server.Just "Full Managed". So they will keep your server up2date. And manage your server, install things for you etc. That are my requirements if I will choose for a Managed Dedicated.But I don't have managed dedicated for my self
Its a good option to select a perfect package for your server and then signup.. that way you know what you are going to get and what you want to be done on your server.