Bugzilla

:jmaher :jgilbert

Azure Virtual Machines allow for different types of disk layout. More information about each type of disk [here.](https://learn.microsoft.com/en-us/azure/virtual-machines/managed-disks-overview#disk-roles) We have migrated from using managed disks to using ephemeral disks to reduce cost and increase performance.

Prior to March 2024, each Azure Virtual Machine was using the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: A managed disk that had 2 partitions, Z:\ and Y:\. Z:\ is where `tasksDir` for Generic Worker was set and Y:\ was where each task stored cached hg and pip packages.
**Temporary Disk**: Located on D:\, which was only used to store pagefile.

After March 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

After June 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which is no longer a managed disk, but rather an ephemeral OS disk that has base software installed.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

Prior to March 2024, we were setting specific worker pools to have different skus on the data disk (more info [here](https://learn.microsoft.com/en-us/azure/virtual-machines/disks-types#disk-type-comparison) based on the worker pool. For example, worker pools with `-ssd` or `-source` were often set to Premium SSD SKU.

We now have it so that the worker pools named `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` are all running ephemeral OS disk to strike a balance between cost and performance.

With all of that said, I would imagine we'd see a better improvement with the way `tarfile.extractall()` now. :jgilbert would you be able to validate that all of these data disk adjustments in the last few months have improved performance?

:jmaher :jgilbert

Azure Virtual Machines allow for different types of disk layout. More information about each type of disk [here.](https://learn.microsoft.com/en-us/azure/virtual-machines/managed-disks-overview#disk-roles) We have migrated from using managed disks to using ephemeral disks to reduce cost and increase performance.

Prior to March 2024, each Azure Virtual Machine was using the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: A managed disk that had 2 partitions, Z:\ and Y:\. Z:\ is where `tasksDir` for Generic Worker was set and Y:\ was where each task stored cached hg and pip packages.
**Temporary Disk**: Located on D:\, which was only used to store pagefile.

After March 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

After June 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which is no longer a managed disk, but rather an ephemeral OS disk that has base software installed.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

Prior to March 2024, we were setting specific worker pools to have different skus on the data disk (more info [here](https://learn.microsoft.com/en-us/azure/virtual-machines/disks-types#disk-type-comparison) based on the worker pool. For example, worker pools with `-ssd` or `-source` were often set to Premium SSD SKU.

We now have it so that the worker pools named `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` are all running ephemeral OS disks to strike a balance between cost and performance.

With all of that said, I would imagine we'd see a better improvement with the way `tarfile.extractall()` now. :jgilbert would you be able to validate that all of these data disk adjustments in the last few months have improved performance?

:jmaher :jgilbert

Azure Virtual Machines allow for different types of disk layout. More information about each type of disk [here.](https://learn.microsoft.com/en-us/azure/virtual-machines/managed-disks-overview#disk-roles) We have migrated from using managed disks to using ephemeral disks to reduce cost and increase performance.

Prior to March 2024, each Azure Virtual Machine was using the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: A managed disk that had 2 partitions, Z:\ and Y:\. Z:\ is where `tasksDir` for Generic Worker was set and Y:\ was where each task stored cached hg and pip packages.
**Temporary Disk**: Located on D:\, which was only used to store pagefile.

After March 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

After June 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which is no longer a managed disk, but rather an ephemeral OS disk that has base software installed.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

Prior to March 2024, we were setting specific worker pools to have different skus on the data disk (more info [here](https://learn.microsoft.com/en-us/azure/virtual-machines/disks-types#disk-type-comparison) based on the worker pool. For example, worker pools with `-ssd` or `-source` were often set to Premium SSD SKU.

We now have it so that the worker pools named `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` are all running ephemeral OS disks to strike a balance between cost and performance.

With all of that said, I would imagine we'd see a better improvement `tarfile.extractall()`. :jgilbert would you be able to validate that all of these data disk adjustments in the last few months have improved performance?

:jmaher :jgilbert

Azure Virtual Machines allow for different types of disk layout. More information about each type of disk [here.](https://learn.microsoft.com/en-us/azure/virtual-machines/managed-disks-overview#disk-roles) We have migrated from using managed disks to using ephemeral disks to reduce cost and increase performance.

Prior to March 2024, each Azure Virtual Machine was using the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: A managed disk that had 2 partitions, Z:\ and Y:\. Z:\ is where `tasksDir` for Generic Worker was set and Y:\ was where each task stored cached hg and pip packages.
**Temporary Disk**: Located on D:\, which was only used to store pagefile.

After March 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which was a managed disk where we installed base software.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

After June 2024, we updated the `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` pools with the following for data storage:

**OS Disk**: Located at C:\, which is no longer a managed disk, but rather an ephemeral OS disk that has base software installed.
**Data Disk**: No longer used.
**Temporary Disk**: Located on D:\ which is where `tasksDir` is set to, along with cached hg and pip packages.

Prior to March 2024, we were setting specific worker pools to have different skus on the data disk (more info [here](https://learn.microsoft.com/en-us/azure/virtual-machines/disks-types#disk-type-comparison) based on the worker pool). For example, worker pools with `-ssd` or `-source` were often set to Premium SSD SKU.

We now have it so that the worker pools named `gecko-t/win10*`,`gecko-t/win11*`, and `gecko-t/win2022*` are all running ephemeral OS disks to strike a balance between cost and performance.

With all of that said, I would imagine we'd see a better improvement `tarfile.extractall()`. :jgilbert would you be able to validate that all of these data disk adjustments in the last few months have improved performance?