Veeam Agent for Linux is now available!
-
@BRRABill said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@JaredBusch said in Veeam Agent for Linux is now available!:
No one with a brain is a fan of agentless full backups. Conveniently, real backup solutions do not need to make constant full backups. Veeam calls it forward incremental-forever, and while I do not know Unitrends' name, I know they do it also.
This is also what I like to do - incrementals.
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
That's why a lot of my initial discussions were on image backups.
As a SMB (or SOHO as @scottalanmiller would say) we have low storage servers here. So in a small amount of time I can get things back up and running as they were. If we had a ton of data it would be a different story altogether.
You assume that image backups are faster. Maybe they are, in some cases, but only barely. Modern DevOps style backups are insanely fast.
-
@Dashrender said in Veeam Agent for Linux is now available!:
I can't speak for JB's response, but my response was in reference to:
@scottalanmiller said in Veeam Agent for Linux is now available!:
I'm not a fan of agentless full backups. I prefer rapid cloning and only restoring needed data.
As I read it, this response was generic, not specifically related to the OP.
Correct
-
@Dashrender said in Veeam Agent for Linux is now available!:
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
Yes, build the OS from a standard template for even greater speed, more repeatable processes and less data to transfer. Unless you have no bottlenecks from the backup storage to the live storage, your way is slower. You have to wait for more to be restored.
-
@scottalanmiller said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
Yes, build the OS from a standard template for even greater speed, more repeatable processes and less data to transfer. Unless you have no bottlenecks from the backup storage to the live storage, your way is slower. You have to wait for more to be restored.
This assumes the configurations can all be done via scripts - granted in Linux based OSes that's almost always the case, but not always in windows.
-
@stacksofplates said in Veeam Agent for Linux is now available!:
@BRRABill said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@JaredBusch said in Veeam Agent for Linux is now available!:
No one with a brain is a fan of agentless full backups. Conveniently, real backup solutions do not need to make constant full backups. Veeam calls it forward incremental-forever, and while I do not know Unitrends' name, I know they do it also.
This is also what I like to do - incrementals.
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
That's why a lot of my initial discussions were on image backups.
As a SMB (or SOHO as @scottalanmiller would say) we have low storage servers here. So in a small amount of time I can get things back up and running as they were. If we had a ton of data it would be a different story altogether.
Linux helps a lot with this. I can either make a template and clone it which takes about 1-2 seconds to finish. I can also run virt-builder and inject ssh keys and scripts to build the image the way I want. I usually do a mix. I make a template, then use virt-sysprep to prepare for cloning and then I can run virt-customize --update --selinux-relabel to update the packages in the image so it's always ready to clone.
So I can have a new full VM up and running in 5-7 seconds (SELinux takes some time to relabel everything, otherwise it would be 1-2 seconds).
Most SMBs don't have storage subsystems that fast, but I get your point.
-
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
@BRRABill said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@JaredBusch said in Veeam Agent for Linux is now available!:
No one with a brain is a fan of agentless full backups. Conveniently, real backup solutions do not need to make constant full backups. Veeam calls it forward incremental-forever, and while I do not know Unitrends' name, I know they do it also.
This is also what I like to do - incrementals.
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
That's why a lot of my initial discussions were on image backups.
As a SMB (or SOHO as @scottalanmiller would say) we have low storage servers here. So in a small amount of time I can get things back up and running as they were. If we had a ton of data it would be a different story altogether.
Linux helps a lot with this. I can either make a template and clone it which takes about 1-2 seconds to finish. I can also run virt-builder and inject ssh keys and scripts to build the image the way I want. I usually do a mix. I make a template, then use virt-sysprep to prepare for cloning and then I can run virt-customize --update --selinux-relabel to update the packages in the image so it's always ready to clone.
So I can have a new full VM up and running in 5-7 seconds (SELinux takes some time to relabel everything, otherwise it would be 1-2 seconds).
Most SMBs don't have storage subsystems that fast, but I get your point.
That's on a DL380 G6 with refurb 300GB 10K SAS. It's fast no matter what you use.
-
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
-
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
-
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
15GB, but it's qcow2 so it's thinprovisioned.
-
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
I have a cronjob that does virt-customize to update packages in the image. So that means I can literally clone this system, and have a full webserver running before you can clone a Windows VM.
-
@Dashrender said in Veeam Agent for Linux is now available!:
@scottalanmiller said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
Yes, build the OS from a standard template for even greater speed, more repeatable processes and less data to transfer. Unless you have no bottlenecks from the backup storage to the live storage, your way is slower. You have to wait for more to be restored.
This assumes the configurations can all be done via scripts - granted in Linux based OSes that's almost always the case, but not always in windows.
It's pretty rare that configuration can't be done via script. What example would you have where it cannot be? Unless you are running Windows NT or DOS. Windows is 100% controllable by PowerShell for some generations. So Windows can't be a factor.
-
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
@BRRABill said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@JaredBusch said in Veeam Agent for Linux is now available!:
No one with a brain is a fan of agentless full backups. Conveniently, real backup solutions do not need to make constant full backups. Veeam calls it forward incremental-forever, and while I do not know Unitrends' name, I know they do it also.
This is also what I like to do - incrementals.
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
That's why a lot of my initial discussions were on image backups.
As a SMB (or SOHO as @scottalanmiller would say) we have low storage servers here. So in a small amount of time I can get things back up and running as they were. If we had a ton of data it would be a different story altogether.
Linux helps a lot with this. I can either make a template and clone it which takes about 1-2 seconds to finish. I can also run virt-builder and inject ssh keys and scripts to build the image the way I want. I usually do a mix. I make a template, then use virt-sysprep to prepare for cloning and then I can run virt-customize --update --selinux-relabel to update the packages in the image so it's always ready to clone.
So I can have a new full VM up and running in 5-7 seconds (SELinux takes some time to relabel everything, otherwise it would be 1-2 seconds).
Most SMBs don't have storage subsystems that fast, but I get your point.
If they don't, you have a problem. How do you have any subsystem that takes longer than that to clone an image?
-
@stacksofplates said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
@BRRABill said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@JaredBusch said in Veeam Agent for Linux is now available!:
No one with a brain is a fan of agentless full backups. Conveniently, real backup solutions do not need to make constant full backups. Veeam calls it forward incremental-forever, and while I do not know Unitrends' name, I know they do it also.
This is also what I like to do - incrementals.
Scott - you're saying that you prefer to completely disregard the OS and only backup the data - OK fine, but this makes full recovery take longer as you have to install an OS, then configure it, then restore the data. In Dev Osp that's fine, but most of use don't have Dev Ops setups. I suppose many workloads could look at moving that direction, even in SMB, I'm just curious about the skills required to manage/maintain it.
That's why a lot of my initial discussions were on image backups.
As a SMB (or SOHO as @scottalanmiller would say) we have low storage servers here. So in a small amount of time I can get things back up and running as they were. If we had a ton of data it would be a different story altogether.
Linux helps a lot with this. I can either make a template and clone it which takes about 1-2 seconds to finish. I can also run virt-builder and inject ssh keys and scripts to build the image the way I want. I usually do a mix. I make a template, then use virt-sysprep to prepare for cloning and then I can run virt-customize --update --selinux-relabel to update the packages in the image so it's always ready to clone.
So I can have a new full VM up and running in 5-7 seconds (SELinux takes some time to relabel everything, otherwise it would be 1-2 seconds).
Most SMBs don't have storage subsystems that fast, but I get your point.
That's on a DL380 G6 with refurb 300GB 10K SAS. It's fast no matter what you use.
RAID 1 WD Reds would be only ten seconds or so.
-
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
How the heck did you clone it?
-
@stacksofplates said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
I have a cronjob that does virt-customize to update packages in the image. So that means I can literally clone this system, and have a full webserver running before you can clone a Windows VM.
I have a base image for different OSes so it's never more than a few seconds.
-
@scottalanmiller said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
I have a cronjob that does virt-customize to update packages in the image. So that means I can literally clone this system, and have a full webserver running before you can clone a Windows VM.
I have a base image for different OSes so it's never more than a few seconds.
I really only use RHEL/CentOS so that's the only image I use. But once I set the hostname and generate/sign the puppet cert it's like 30 seconds or so.
But even if it's a one off I don't have to worry about updating packages after cloning since that runs nightly in the template.
-
@stacksofplates said in Veeam Agent for Linux is now available!:
@scottalanmiller said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
@Dashrender said in Veeam Agent for Linux is now available!:
@stacksofplates said in Veeam Agent for Linux is now available!:
It's a 15 GB qcow2 with only the OS (which is way overkill for single service servers). The speed is coming from separating the data from the OS, not from the storage.
how small is the VM you're cloning?
When I've cloned a 40 GB Windows install, it took more than 1-2 second, more like a min+ not that even a min matters here compared to installing a whole Windows server OS and patches, etc.
I have a cronjob that does virt-customize to update packages in the image. So that means I can literally clone this system, and have a full webserver running before you can clone a Windows VM.
I have a base image for different OSes so it's never more than a few seconds.
I really only use RHEL/CentOS so that's the only image I use. But once I set the hostname and generate/sign the puppet cert it's like 30 seconds or so.
But even if it's a one off I don't have to worry about updating packages after cloning since that runs nightly in the template.
That's all I have in production, too
-
@Dashrender it was 2-3 seconds to clone, then the normal boot time. A little faster at work with recent hardware. It's quicker for me to type it than to click through the clone menu and rename VM and drive.
-
Anyone actually running Veeam Agent for Linux as their backup?
-
@scottalanmiller said in Veeam Agent for Linux is now available!:
Anyone actually running Veeam Agent for Linux as their backup?
I'm still on the last beta for a couple of machines. It works surprisingly well.
