[Users] Limit swap space on ovirt node install
Hi, We have a 16GB SSD in all our nodes to install the ovirt node image on. While installing I get the message there is not enough space. After checking the requirements of ovirt node it should fit quite easily, but I think the swap space assigned is too much as we have a lot of ram in the nodes. As we don't overcommit the RAM in our nodes we shouldn't need that much swap. Is there any way to modify it? I'd hate to have to buy new SSD's :-( -- Regards, Netbulae Jorick Astrego Netbulae B.V.
You can do this with 2 different methods, both of which are on the kernel command line during boot. Simply hit TAB when you're selecting the install option and add one of the options below to the command line. 1. You can set " storage_vol=:XXX:::: " where XXX is the desired size of the SWAP partition in MB. It has a minimum size of 5MB. This lets you set a static size for swap. 2. You can set mem_overcommit=0. This will change our calculation to not add any swap for overcommit. It will still calculate a value based on the amount of RAM, but not as large as with the overcommit. The swap size is calculated at mem_overcommit*RAM + <BASE_SWAP> BASE_SWAP is calculated using this formula: if MEM_SIZE < 4GB: BASE_SWAP = 2GB elif MEM_SIZE < 16GB: BASE_SWAP = 4GB elif MEM_SIZE < 64GB: BASE_SWAP = 8GB else: BASE_SWAP = 16GB We are planning to expose these option in the TUI installer, just not there yet. Mike ----- Original Message -----
Hi,
We have a 16GB SSD in all our nodes to install the ovirt node image on. While installing I get the message there is not enough space.
After checking the requirements of ovirt node it should fit quite easily, but I think the swap space assigned is too much as we have a lot of ram in the nodes.
As we don't overcommit the RAM in our nodes we shouldn't need that much swap. Is there any way to modify it? I'd hate to have to buy new SSD's :-(
-- Regards,
Netbulae Jorick Astrego
Netbulae B.V.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Hi Mark, It took a while to have the time to continue testing. I tried both methods. When I use " storage_vol=:4000:::: " and " mem_overcommit=0" everything installs fine. But after I reboot the node i get the screen "login as admin to configure" and the system hangs. After several seconds I get the message "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system" When I only use " mem_overcommit=0 ", I cannot install the node as the Disk Space Check tells me I need a bigger drive. "Available Space: 14305MB" "Required Space: 26782MB" This should be less: 0*32 + 8(32<64) = 8GB Swap. Plus the storage ovirt needs itself (300MB?) I tried ovirt-node 2.5.1 and 2.5.2 Kind regards, Jorick Astrego Netbulae B.V. On 09/01/2012 10:34 PM, Mike Burns wrote:
You can do this with 2 different methods, both of which are on the kernel command line during boot. Simply hit TAB when you're selecting the install option and add one of the options below to the command line.
1. You can set " storage_vol=:XXX:::: " where XXX is the desired size of the SWAP partition in MB. It has a minimum size of 5MB. This lets you set a static size for swap.
2. You can set mem_overcommit=0. This will change our calculation to not add any swap for overcommit. It will still calculate a value based on the amount of RAM, but not as large as with the overcommit.
The swap size is calculated at mem_overcommit*RAM + <BASE_SWAP>
BASE_SWAP is calculated using this formula:
if MEM_SIZE < 4GB: BASE_SWAP = 2GB elif MEM_SIZE < 16GB: BASE_SWAP = 4GB elif MEM_SIZE < 64GB: BASE_SWAP = 8GB else: BASE_SWAP = 16GB
We are planning to expose these option in the TUI installer, just not there yet.
Mike
----- Original Message -----
Hi,
We have a 16GB SSD in all our nodes to install the ovirt node image on. While installing I get the message there is not enough space.
After checking the requirements of ovirt node it should fit quite easily, but I think the swap space assigned is too much as we have a lot of ram in the nodes.
As we don't overcommit the RAM in our nodes we shouldn't need that much swap. Is there any way to modify it? I'd hate to have to buy new SSD's :-(
-- Regards,
Netbulae Jorick Astrego
Netbulae B.V.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
-- Met vriendelijke groet, Netbulae Jorick Astrego Netbulae B.V. Staalsteden 4-13 7547 TA Enschede Tel. +31 (0)53 - 20 30 270 Email: jorick@netbulae.eu Site: http://www.netbulae.eu
Hi, Finally I found the problem but I still don't understand it. The problem appears to be that I install using the Supermicro IPMI with local cdrom. Installation goes well but after that I am unable to login remotely and get the error "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system". The system appears to hang. When I hook up a screen and keyboard I can login normally and everything works fine, this is very strange and annoying as I now need to install every node in the datacenter. I also tried logging in after startup but when I check the IPMI, I only get a black screen. Now I started using cobbler for using PXE boot with latest nightly release, but I still have to install and configure the node. How can I push the installation automatically so I can install unattended? Any ideas? Kind regards, Jorick Astrego On 10/15/2012 03:57 PM, Jorick Astrego wrote:
Hi Mark,
It took a while to have the time to continue testing.
I tried both methods. When I use " storage_vol=:4000:::: " and " mem_overcommit=0" everything installs fine.
But after I reboot the node i get the screen "login as admin to configure" and the system hangs.
After several seconds I get the message "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system"
When I only use " mem_overcommit=0 ", I cannot install the node as the Disk Space Check tells me I need a bigger drive.
"Available Space: 14305MB" "Required Space: 26782MB"
This should be less:
0*32 + 8(32<64) = 8GB Swap.
Plus the storage ovirt needs itself (300MB?)
I tried ovirt-node 2.5.1 and 2.5.2
Kind regards,
Jorick Astrego Netbulae B.V.
On 09/01/2012 10:34 PM, Mike Burns wrote:
You can do this with 2 different methods, both of which are on the kernel command line during boot. Simply hit TAB when you're selecting the install option and add one of the options below to the command line.
1. You can set " storage_vol=:XXX:::: " where XXX is the desired size of the SWAP partition in MB. It has a minimum size of 5MB. This lets you set a static size for swap.
2. You can set mem_overcommit=0. This will change our calculation to not add any swap for overcommit. It will still calculate a value based on the amount of RAM, but not as large as with the overcommit.
The swap size is calculated at mem_overcommit*RAM + <BASE_SWAP>
BASE_SWAP is calculated using this formula:
if MEM_SIZE < 4GB: BASE_SWAP = 2GB elif MEM_SIZE < 16GB: BASE_SWAP = 4GB elif MEM_SIZE < 64GB: BASE_SWAP = 8GB else: BASE_SWAP = 16GB
We are planning to expose these option in the TUI installer, just not there yet.
Mike
----- Original Message -----
Hi,
We have a 16GB SSD in all our nodes to install the ovirt node image on. While installing I get the message there is not enough space.
After checking the requirements of ovirt node it should fit quite easily, but I think the swap space assigned is too much as we have a lot of ram in the nodes.
As we don't overcommit the RAM in our nodes we shouldn't need that much swap. Is there any way to modify it? I'd hate to have to buy new SSD's :-(
-- Regards,
Netbulae Jorick Astrego
Netbulae B.V.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
On Fri, 2012-10-19 at 14:26 +0200, Jorick Astrego wrote:
Hi,
Finally I found the problem but I still don't understand it.
The problem appears to be that I install using the Supermicro IPMI with local cdrom.
Installation goes well but after that I am unable to login remotely and get the error "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system". The system appears to hang.
When I hook up a screen and keyboard I can login normally and everything works fine, this is very strange and annoying as I now need to install every node in the datacenter. I also tried logging in after startup but when I check the IPMI, I only get a black screen.
Now I started using cobbler for using PXE boot with latest nightly release, but I still have to install and configure the node. How can I push the installation automatically so I can install unattended?
We apparently don't have the options documented on ovirt.org anywhere, but I'll make sure we do that. In the meantime, you can refer to this document [1] for automated installation options. The options are pretty much identical. Thanks Mike [1] http://goo.gl/vbeXQ
Any ideas?
Kind regards,
Jorick Astrego
On 10/15/2012 03:57 PM, Jorick Astrego wrote:
Hi Mark,
It took a while to have the time to continue testing.
I tried both methods. When I use " storage_vol=:4000:::: " and " mem_overcommit=0" everything installs fine.
But after I reboot the node i get the screen "login as admin to configure" and the system hangs.
After several seconds I get the message "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system"
When I only use " mem_overcommit=0 ", I cannot install the node as the Disk Space Check tells me I need a bigger drive.
"Available Space: 14305MB" "Required Space: 26782MB"
This should be less:
0*32 + 8(32<64) = 8GB Swap.
Plus the storage ovirt needs itself (300MB?)
I tried ovirt-node 2.5.1 and 2.5.2
Kind regards,
Jorick Astrego Netbulae B.V.
On 09/01/2012 10:34 PM, Mike Burns wrote:
You can do this with 2 different methods, both of which are on the kernel command line during boot. Simply hit TAB when you're selecting the install option and add one of the options below to the command line.
1. You can set " storage_vol=:XXX:::: " where XXX is the desired size of the SWAP partition in MB. It has a minimum size of 5MB. This lets you set a static size for swap.
2. You can set mem_overcommit=0. This will change our calculation to not add any swap for overcommit. It will still calculate a value based on the amount of RAM, but not as large as with the overcommit.
The swap size is calculated at mem_overcommit*RAM + <BASE_SWAP>
BASE_SWAP is calculated using this formula:
if MEM_SIZE < 4GB: BASE_SWAP = 2GB elif MEM_SIZE < 16GB: BASE_SWAP = 4GB elif MEM_SIZE < 64GB: BASE_SWAP = 8GB else: BASE_SWAP = 16GB
We are planning to expose these option in the TUI installer, just not there yet.
Mike
----- Original Message -----
Hi,
We have a 16GB SSD in all our nodes to install the ovirt node image on. While installing I get the message there is not enough space.
After checking the requirements of ovirt node it should fit quite easily, but I think the swap space assigned is too much as we have a lot of ram in the nodes.
As we don't overcommit the RAM in our nodes we shouldn't need that much swap. Is there any way to modify it? I'd hate to have to buy new SSD's :-(
-- Regards,
Netbulae Jorick Astrego
Netbulae B.V.
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
On 10/22/2012 02:26 PM, Mike Burns wrote:
On Fri, 2012-10-19 at 14:26 +0200, Jorick Astrego wrote:
Hi,
Finally I found the problem but I still don't understand it.
The problem appears to be that I install using the Supermicro IPMI with local cdrom.
Installation goes well but after that I am unable to login remotely and get the error "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system". The system appears to hang.
When I hook up a screen and keyboard I can login normally and everything works fine, this is very strange and annoying as I now need to install every node in the datacenter. I also tried logging in after startup but when I check the IPMI, I only get a black screen.
Now I started using cobbler for using PXE boot with latest nightly release, but I still have to install and configure the node. How can I push the installation automatically so I can install unattended? We apparently don't have the options documented on ovirt.org anywhere, but I'll make sure we do that. In the meantime, you can refer to this document [1] for automated installation options. The options are pretty much identical.
Thanks
Mike
Mike, Thanks for the answer, I'm deploying the latest version as we speak. This will save so much time. But it is still strange I cannot use the IPMI, is there any info I can supply for this? Should I enter it in Bugzilla? Thanks for all the help, Jorick
On Tue, 2012-10-23 at 09:30 +0200, Jorick Astrego wrote:
On 10/22/2012 02:26 PM, Mike Burns wrote:
On Fri, 2012-10-19 at 14:26 +0200, Jorick Astrego wrote:
Hi,
Finally I found the problem but I still don't understand it.
The problem appears to be that I install using the Supermicro IPMI with local cdrom.
Installation goes well but after that I am unable to login remotely and get the error "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system". The system appears to hang.
When I hook up a screen and keyboard I can login normally and everything works fine, this is very strange and annoying as I now need to install every node in the datacenter. I also tried logging in after startup but when I check the IPMI, I only get a black screen.
Now I started using cobbler for using PXE boot with latest nightly release, but I still have to install and configure the node. How can I push the installation automatically so I can install unattended? We apparently don't have the options documented on ovirt.org anywhere, but I'll make sure we do that. In the meantime, you can refer to this document [1] for automated installation options. The options are pretty much identical.
Thanks
Mike
Mike,
Thanks for the answer, I'm deploying the latest version as we speak. This will save so much time.
But it is still strange I cannot use the IPMI, is there any info I can supply for this? Should I enter it in Bugzilla?
We don't do anything special with ipmi currently. What type of system and what type of management device is being used? Mike
Thanks for all the help,
Jorick
On 10/23/2012 01:58 PM, Mike Burns wrote:
On Tue, 2012-10-23 at 09:30 +0200, Jorick Astrego wrote:
On 10/22/2012 02:26 PM, Mike Burns wrote:
On Fri, 2012-10-19 at 14:26 +0200, Jorick Astrego wrote:
Hi,
Finally I found the problem but I still don't understand it.
The problem appears to be that I install using the Supermicro IPMI with local cdrom.
Installation goes well but after that I am unable to login remotely and get the error "systemd-readahead-collect[500]: Failed to open pack file: Read-only file system". The system appears to hang.
When I hook up a screen and keyboard I can login normally and everything works fine, this is very strange and annoying as I now need to install every node in the datacenter. I also tried logging in after startup but when I check the IPMI, I only get a black screen.
Now I started using cobbler for using PXE boot with latest nightly release, but I still have to install and configure the node. How can I push the installation automatically so I can install unattended? We apparently don't have the options documented on ovirt.org anywhere, but I'll make sure we do that. In the meantime, you can refer to this document [1] for automated installation options. The options are pretty much identical.
Thanks
Mike
Mike,
Thanks for the answer, I'm deploying the latest version as we speak. This will save so much time.
But it is still strange I cannot use the IPMI, is there any info I can supply for this? Should I enter it in Bugzilla? We don't do anything special with ipmi currently. What type of system and what type of management device is being used?
Mike That's what I find so strange, installing works but configuring hangs.
I use Supermicro onboard IPMI 2.0 (http://www.supermicro.com/products/nfo/IPMI.cfm). Could it be something to do with USB redirection, I've had problems before with the Virtual CD but that has been fixed in the nightly build. Jorick
participants (2)
-
Jorick Astrego -
Mike Burns