Cloudstack: FAQ
Posted: Fri Jul 12, 2013 3:42 am
Hello !
Due to our cloud services being considerably different than the standard VPS, they need a special FAQ section where mostly the differences will be pointed out.
For the other aspects, please look at the main FAQ instead.
Q0: What is a cloud and when do I need cloud services ?
A0. As of late, the term cloud has been abused, many providers claiming they offer cloud services when in fact this is not the case.
While there is not a scientific definition of the term "cloud computing" in my view it has to:
-Scale fast and almost unlimited for both the users and the host;
-Have HA and fail-over;
-Allow complex infrastructure allocation (i.e. internal routing, switching, bridging, firewalling, etc) while maintaining strict separation of customers/infrastructures (VLANs, firewalling, dedicated resources).
By pooling together many servers, big and flexible storage, networking, etc, economies of scale can be done, not only in utilization of the resources (less wasted space, CPU cycles, RAM, etc), but also in hardware. However, since, in my view, redundancy and HA are absolutely needed for the service to be called Cloud, there is a significant overhead due to some resources being always on stand-by in order to be available in case of hardware failure. also, a lot of resources will be unused to give people the ability to grow whenever they suddenly need it.
Overall, cloud computing needs much more knowledge and experience in administration (for the host), not only technical expertise, but also mathematics and economy knowledge in order to make an efficient allocation of resources, without hitting the ceiling or floor too many times, even in unexpected events such as serious hardware failure.
Based on the above description you will need cloud services when you:
-Need fast deployment of a VM, based on a template or disk image you did and you only need it for a while, say a few hours or a few days;
-Run important services and you lose a lot when these are down, so you need to be shielded from hardware failure/extended downtime;
-Need complex infrastructures at a short notice, such as a virtual office for a new division/aquired company, you can simply rent some IaaS and partition it as you like;
-You frown upon hardware costs and fast deprecation and wish to have a good stable infrastructure without having to hire someone to set up the hardware and pay for it or you only needed for a few months and dont want to waste time and money buying second hand equipment and then having to sell it, for example.
There are many usage scenarios, a real cloud will be able to do all those things, while, what is marketed as cloud by many others will not. We aim to have a real cloud infrastructure, with ALL the features, giving you full control (own iso, upload/download snapshots and drive images, create/destroy instances and VLANs at will, etc) It is also a great learning tool to test some deployments before actually buying the hardware, for example. You can have many "machines" and complex "switches/routers" simulated in just a couple of hours of work, most of it needed to setup the OSes and applications, not for creating the infrastructure which is mostly point and click.
Q1 What is HA ?
A1. HA stands for High Availability and consists of a watchdog which checks the state of your VM to always be powered up. Just try to shutdown your VM from inside, you will see it restarting almost immediately it went down.
In my view, HA includes failover, i.e. if your VM is down due to hardware failure, the instance will be restarted on another node and you will notice just a minute or 2 of downtime, however, many people do not agree, for most HA is HA, failover is failover, even if I think HA will not help in case of hardware failure if failover is not implemented.
Q2. I want to add IPv6 to my instance ! Where can I find the allocated IPv6 and instructions on how to do this ?
A2. Due to the feedback so far and for other reasons, we opted to disable automated IPv6 configuration based on MAC (SLAAC), however, this does not mean your VPS will not be able to use IPv6.
Your allocated /64 can be found in relation to your current IP. Just access this link: http://prometeus.net/returnipv6.php?IP=xxx.xxx.xxx.xxx (replace xxx.xxx.xxx.xxx with your IPv4).
As for how to add it in your configuration files (or in a gui for desktops systems), that depends on distribution.
CentOS (and CentOS/RHEL based): http://www.cyberciti.biz/faq/rhel-redha ... iguration/
Debian/Ubuntu (and Debian/Ubuntu based): http://www.cyberciti.biz/faq/debian-lin ... iguration/
You can use any of the IPs in the /64, or more of them at once, including the whole range, however, they are 18,446,744,073,709,551,616 please consider there is not enough memory to load them all, especially if you have a firewall in place for it too (the security zone outside the instance will only work for IPv4, therefore you need to use an IPv6 firewall inside the VMs if you wish to restrict access to IPv6).
Due to a limitation in CloudStack, for now, the virtual router does not support IPv6, therefore only shared networks can use it while the isolated ones cannot.
Q3. I have just destroyed an recreated my instance with my own template/custom config network and it does not work anymore !
A3. This is possibly due to the fact you configured static IPV4/IPv6 and by destroying and re-creating the instance, you have got another IP set. Please go to the console and issue the command in order to find which is your new IPv4 and derive from it the IPv6 /64 as shown above. The firewall will not need to be redone, but you will have to change the IPs in order to make the networking work. If you wish to preserve your IP, the only sure way is to keep the instance (even powered off will keep the IP, just don't destroy it).
Q4. I need rDNS/PTR records ! Where can I add them to my IP(s) ?
A4. The cloud services are meant for creating/destroying instances as you need them, as well as for long term hosting with HA/failover. In the first case, you will keep changing the IP, so maintaining rDNS/PTR records will be a nightmare, in the second case, this is fully justified and expected, but so far, CloudStack does not have this possibility and our custom panel is not ready, but it will probably be implemented.
For now, you will need to open a ticket, but only if you need long time hosting with HA, not if you plan to keep creating/destroying instances. In the latter case, please use a SMTP service or a relay you setup some other place.
Q5. I wish to grow/shrink my primary storage (ROOT or DATA disks) How can I do that ?
A5. This is possible through the API, however, simply growing the disk is not enough, more disk space will be added but you still need to enlarge the partition. The recommended way is to add disks and format/mount them from inside the VM. You can also move them around to attach to another VM or even clone them once you have there the data needed to another VM too.
Here is the API for resizing volumes: https://cwiki.apache.org/CLOUDSTACK/res ... lumes.html
Please also take a look at the CloudMonkey API tutorial here: viewtopic.php?f=8&t=1360
Q6. I wish to prepay for x months ! You only have one-time payments, why is that ?
A6. This is not how it works. Think of it as to a prepaid sim card where the credit never expires. You can buy it once, do some calls, when the credit is 0 you can no longer use it, but you can recharge it. It is up to you how many calls you do and how long will they be (create and use VMs), you do not need to pay a subscription, you just reload credits when your current credit is done (or before that, it is all up to you). Therefore, there are no schedules anywhere, you cannot pay x months of service because the service is not fixed, you can at any time create more instances, shut down some, even destroy them, add more snapshots, ISOs, templates, etc.
Q7. How is the billing done ? Am I being billed if the VM is shut down ?
A7. You are being billed only for used resources. For example, if your VM is up and running you pay for:
1. The instance itself (RAM, CPU(s), Traffic, IP(s));
2. Main storage (DATA and ROOT disk images);
3. Secondary storage (Templates, snapshots, ISOs);
4. The virtual router if you are running an isolated network.
If you shut down the VM, then you no longer pay for the instance as it is no longer using RAM, CPU and traffic, but you do pay for the IP because it is reserved, we can't reuse it. All the other resources (Storage, virtual router if it applies) are still in use and you will continue to pay them.
If you also destroy the instance, for example, you only wish to have a template ready to be deployed when needed, you no longer pay the instance, the main storage and the IP(s) (they are released back into the pool and next time you create an instance you will get others), you continue to pay the secondary storage (for the template space).
Q8. You say I can mount my own ISO. How do I do that ?
A8. Please go to Templates, select ISO from the drop-down menu, then Register ISO in the upper right corner. You can only use HTTP or HTTPs links (ports 80 and 443). Once you give it the command to load, it goes on in the background, it will take some time until the ISO is available depending on how fast is the mirror with the ISO. If really slow, it may even time out, please use good/close mirrors for this.
Also, please select the correct OS when you are registering the ISO, if your release is not listed yet, please select the closest available, for example, if your ISO is Fedora 19 (we have it already loaded, but just as an example), use fedora 13, the last version there is a template for. If your OS is not listed, please use Other PV 64 or 32. Failure to do so will probably create a vm without paravirtualization and the performance will be significantly degraded.
The ISO will then appear under the My ISOs tab for selection when you create a VM from the ISO.
Q9. Which OSes are supported ?
In theory, all. You need, however, to select the correct template (see Q8). If you can install from ISO, then chances are it is supported. Please try to use virtio (Other PV 64 or 32) for best performance. We do not support running VMs without Virtio, but they should work, albeit, slower.
Q10. This cloud stuff is new to me ! Where can I get more information/tutorials ?
Please take a look at this forum, the Cloud tutorials are marked with "Cloudstack". A selection is also available here: http://www.iwstack.com/tutorials/
For deeper understanding, please take a look here: https://cwiki.apache.org/confluence/dis ... STACK/Home
Q11. The prices are really low. X company offers cloud services at much higher prices. What's the catch ?
One of the most important things is that we are only using open source software for the implementation of our cloud. Many providers use VMWare or OnApp, to mention just a few and the licensing costs are very high. In turn, we worked hard to implement our solution, we did not use a ready-made one and the costs for licenses is 0.
Another thing is that we do not have a lot of overhead, we do not employ hundreds of people, we are a small family business with a few thousands customers only.
We also own all hardware and dont have to pay rent for it, while also do not have to answer some investors and turn a big profit. This is a long term operation, we will continuously develop the cloud and add new features such as own control panel for it.
You may also notice that many of the "cloud" companies do not really offer the cloud features too. Some don't even have snapshots, not to mention full HA and fail-over and still charge the same or more than us.
Q12. What is the uptime guarantee ?
A12. While we do not offer a SLA, that is not because we do not manage to keep it, but because our customers already know how reliable we are.
Regular VMs on isolated nodes with local storage have hundreds of days and even more than a year uptime, the network was never down except a few minutes when switches are changed, however, we did have some hardware failures, like everyone.
With this setup, we have full HA and fail-over so nodes can be taken offline for maintenance without interruption of service or can simply fail and the VMs will restart on other nodes so the uptime will be even better. Even if the master node will fail, the zones will continue to work as before, only that you cannot create/destroy/shutdown, register ISO etc as long as the master is down. We also have A+B separate power lines on all equipment.
We have 20 gbps redundant network with extra spare capacity (non-commit lines to kick in in case of failure or large scale DDoSes), the nodes are linked to more than a switch, in short, we offer a lot of redundancy, failure is unlikely to occur, but we cannot rule out earthquakes, fires, catastrophic power failure. If it never happened in 16 years since we are in business, does not mean it cant happen, just that the chances are very low.
Q13. Do you have a money back guarantee ?
A13 Yes, we do, from now on, we will be processing refunds minus a 1 Eur fee for each for paypal and processing fees. Of course, the money already used will not be returned, just the credit balance without the included promotion, if any.
Due to our cloud services being considerably different than the standard VPS, they need a special FAQ section where mostly the differences will be pointed out.
For the other aspects, please look at the main FAQ instead.
Q0: What is a cloud and when do I need cloud services ?
A0. As of late, the term cloud has been abused, many providers claiming they offer cloud services when in fact this is not the case.
While there is not a scientific definition of the term "cloud computing" in my view it has to:
-Scale fast and almost unlimited for both the users and the host;
-Have HA and fail-over;
-Allow complex infrastructure allocation (i.e. internal routing, switching, bridging, firewalling, etc) while maintaining strict separation of customers/infrastructures (VLANs, firewalling, dedicated resources).
By pooling together many servers, big and flexible storage, networking, etc, economies of scale can be done, not only in utilization of the resources (less wasted space, CPU cycles, RAM, etc), but also in hardware. However, since, in my view, redundancy and HA are absolutely needed for the service to be called Cloud, there is a significant overhead due to some resources being always on stand-by in order to be available in case of hardware failure. also, a lot of resources will be unused to give people the ability to grow whenever they suddenly need it.
Overall, cloud computing needs much more knowledge and experience in administration (for the host), not only technical expertise, but also mathematics and economy knowledge in order to make an efficient allocation of resources, without hitting the ceiling or floor too many times, even in unexpected events such as serious hardware failure.
Based on the above description you will need cloud services when you:
-Need fast deployment of a VM, based on a template or disk image you did and you only need it for a while, say a few hours or a few days;
-Run important services and you lose a lot when these are down, so you need to be shielded from hardware failure/extended downtime;
-Need complex infrastructures at a short notice, such as a virtual office for a new division/aquired company, you can simply rent some IaaS and partition it as you like;
-You frown upon hardware costs and fast deprecation and wish to have a good stable infrastructure without having to hire someone to set up the hardware and pay for it or you only needed for a few months and dont want to waste time and money buying second hand equipment and then having to sell it, for example.
There are many usage scenarios, a real cloud will be able to do all those things, while, what is marketed as cloud by many others will not. We aim to have a real cloud infrastructure, with ALL the features, giving you full control (own iso, upload/download snapshots and drive images, create/destroy instances and VLANs at will, etc) It is also a great learning tool to test some deployments before actually buying the hardware, for example. You can have many "machines" and complex "switches/routers" simulated in just a couple of hours of work, most of it needed to setup the OSes and applications, not for creating the infrastructure which is mostly point and click.
Q1 What is HA ?
A1. HA stands for High Availability and consists of a watchdog which checks the state of your VM to always be powered up. Just try to shutdown your VM from inside, you will see it restarting almost immediately it went down.
In my view, HA includes failover, i.e. if your VM is down due to hardware failure, the instance will be restarted on another node and you will notice just a minute or 2 of downtime, however, many people do not agree, for most HA is HA, failover is failover, even if I think HA will not help in case of hardware failure if failover is not implemented.
Q2. I want to add IPv6 to my instance ! Where can I find the allocated IPv6 and instructions on how to do this ?
A2. Due to the feedback so far and for other reasons, we opted to disable automated IPv6 configuration based on MAC (SLAAC), however, this does not mean your VPS will not be able to use IPv6.
Your allocated /64 can be found in relation to your current IP. Just access this link: http://prometeus.net/returnipv6.php?IP=xxx.xxx.xxx.xxx (replace xxx.xxx.xxx.xxx with your IPv4).
As for how to add it in your configuration files (or in a gui for desktops systems), that depends on distribution.
CentOS (and CentOS/RHEL based): http://www.cyberciti.biz/faq/rhel-redha ... iguration/
Debian/Ubuntu (and Debian/Ubuntu based): http://www.cyberciti.biz/faq/debian-lin ... iguration/
You can use any of the IPs in the /64, or more of them at once, including the whole range, however, they are 18,446,744,073,709,551,616 please consider there is not enough memory to load them all, especially if you have a firewall in place for it too (the security zone outside the instance will only work for IPv4, therefore you need to use an IPv6 firewall inside the VMs if you wish to restrict access to IPv6).
Due to a limitation in CloudStack, for now, the virtual router does not support IPv6, therefore only shared networks can use it while the isolated ones cannot.
Q3. I have just destroyed an recreated my instance with my own template/custom config network and it does not work anymore !
A3. This is possibly due to the fact you configured static IPV4/IPv6 and by destroying and re-creating the instance, you have got another IP set. Please go to the console and issue the
Code: Select all
ifconfigQ4. I need rDNS/PTR records ! Where can I add them to my IP(s) ?
A4. The cloud services are meant for creating/destroying instances as you need them, as well as for long term hosting with HA/failover. In the first case, you will keep changing the IP, so maintaining rDNS/PTR records will be a nightmare, in the second case, this is fully justified and expected, but so far, CloudStack does not have this possibility and our custom panel is not ready, but it will probably be implemented.
For now, you will need to open a ticket, but only if you need long time hosting with HA, not if you plan to keep creating/destroying instances. In the latter case, please use a SMTP service or a relay you setup some other place.
Q5. I wish to grow/shrink my primary storage (ROOT or DATA disks) How can I do that ?
A5. This is possible through the API, however, simply growing the disk is not enough, more disk space will be added but you still need to enlarge the partition. The recommended way is to add disks and format/mount them from inside the VM. You can also move them around to attach to another VM or even clone them once you have there the data needed to another VM too.
Here is the API for resizing volumes: https://cwiki.apache.org/CLOUDSTACK/res ... lumes.html
Please also take a look at the CloudMonkey API tutorial here: viewtopic.php?f=8&t=1360
Q6. I wish to prepay for x months ! You only have one-time payments, why is that ?
A6. This is not how it works. Think of it as to a prepaid sim card where the credit never expires. You can buy it once, do some calls, when the credit is 0 you can no longer use it, but you can recharge it. It is up to you how many calls you do and how long will they be (create and use VMs), you do not need to pay a subscription, you just reload credits when your current credit is done (or before that, it is all up to you). Therefore, there are no schedules anywhere, you cannot pay x months of service because the service is not fixed, you can at any time create more instances, shut down some, even destroy them, add more snapshots, ISOs, templates, etc.
Q7. How is the billing done ? Am I being billed if the VM is shut down ?
A7. You are being billed only for used resources. For example, if your VM is up and running you pay for:
1. The instance itself (RAM, CPU(s), Traffic, IP(s));
2. Main storage (DATA and ROOT disk images);
3. Secondary storage (Templates, snapshots, ISOs);
4. The virtual router if you are running an isolated network.
If you shut down the VM, then you no longer pay for the instance as it is no longer using RAM, CPU and traffic, but you do pay for the IP because it is reserved, we can't reuse it. All the other resources (Storage, virtual router if it applies) are still in use and you will continue to pay them.
If you also destroy the instance, for example, you only wish to have a template ready to be deployed when needed, you no longer pay the instance, the main storage and the IP(s) (they are released back into the pool and next time you create an instance you will get others), you continue to pay the secondary storage (for the template space).
Q8. You say I can mount my own ISO. How do I do that ?
A8. Please go to Templates, select ISO from the drop-down menu, then Register ISO in the upper right corner. You can only use HTTP or HTTPs links (ports 80 and 443). Once you give it the command to load, it goes on in the background, it will take some time until the ISO is available depending on how fast is the mirror with the ISO. If really slow, it may even time out, please use good/close mirrors for this.
Also, please select the correct OS when you are registering the ISO, if your release is not listed yet, please select the closest available, for example, if your ISO is Fedora 19 (we have it already loaded, but just as an example), use fedora 13, the last version there is a template for. If your OS is not listed, please use Other PV 64 or 32. Failure to do so will probably create a vm without paravirtualization and the performance will be significantly degraded.
The ISO will then appear under the My ISOs tab for selection when you create a VM from the ISO.
Q9. Which OSes are supported ?
In theory, all. You need, however, to select the correct template (see Q8). If you can install from ISO, then chances are it is supported. Please try to use virtio (Other PV 64 or 32) for best performance. We do not support running VMs without Virtio, but they should work, albeit, slower.
Q10. This cloud stuff is new to me ! Where can I get more information/tutorials ?
Please take a look at this forum, the Cloud tutorials are marked with "Cloudstack". A selection is also available here: http://www.iwstack.com/tutorials/
For deeper understanding, please take a look here: https://cwiki.apache.org/confluence/dis ... STACK/Home
Q11. The prices are really low. X company offers cloud services at much higher prices. What's the catch ?
One of the most important things is that we are only using open source software for the implementation of our cloud. Many providers use VMWare or OnApp, to mention just a few and the licensing costs are very high. In turn, we worked hard to implement our solution, we did not use a ready-made one and the costs for licenses is 0.
Another thing is that we do not have a lot of overhead, we do not employ hundreds of people, we are a small family business with a few thousands customers only.
We also own all hardware and dont have to pay rent for it, while also do not have to answer some investors and turn a big profit. This is a long term operation, we will continuously develop the cloud and add new features such as own control panel for it.
You may also notice that many of the "cloud" companies do not really offer the cloud features too. Some don't even have snapshots, not to mention full HA and fail-over and still charge the same or more than us.
Q12. What is the uptime guarantee ?
A12. While we do not offer a SLA, that is not because we do not manage to keep it, but because our customers already know how reliable we are.
Regular VMs on isolated nodes with local storage have hundreds of days and even more than a year uptime, the network was never down except a few minutes when switches are changed, however, we did have some hardware failures, like everyone.
With this setup, we have full HA and fail-over so nodes can be taken offline for maintenance without interruption of service or can simply fail and the VMs will restart on other nodes so the uptime will be even better. Even if the master node will fail, the zones will continue to work as before, only that you cannot create/destroy/shutdown, register ISO etc as long as the master is down. We also have A+B separate power lines on all equipment.
We have 20 gbps redundant network with extra spare capacity (non-commit lines to kick in in case of failure or large scale DDoSes), the nodes are linked to more than a switch, in short, we offer a lot of redundancy, failure is unlikely to occur, but we cannot rule out earthquakes, fires, catastrophic power failure. If it never happened in 16 years since we are in business, does not mean it cant happen, just that the chances are very low.
Q13. Do you have a money back guarantee ?
A13 Yes, we do, from now on, we will be processing refunds minus a 1 Eur fee for each for paypal and processing fees. Of course, the money already used will not be returned, just the credit balance without the included promotion, if any.