Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

When running in an hypervisor with SSD disks, detection is not accurate. #41

Open
guettli opened this issue Feb 28, 2019 · 8 comments
Open

Comments

@guettli
Copy link

guettli commented Feb 28, 2019

I read this in the README:

When running in an hypervisor with SSD disks, detection is not accurate.

I guess you are trying to detect if there are SSD disks and then you have a boolean flag.

Why do you do it this way?

Why not check the performance of the disk and set the values according the this value (float value, not boolean)?

@PitoneMaledetto
Copy link

PitoneMaledetto commented Mar 1, 2019

Hi guetti,
maybe is not related to your question above, but do you use the --ssd flag when running the script?
Regards

@guettli
Copy link
Author

guettli commented Mar 1, 2019

@PitoneMaledetto I know the --ssd flag. But I have great feeling. It is love. It is called "automation". I don't to supply a flag. I love it, if scripts can collect all needed information themselves. And in this case I guess it is not black magic.

@jfcoz
Copy link
Owner

jfcoz commented Mar 1, 2019

@guettli , benching disk would be too agressive, it could have a perfomance impact on a production database and/or report bad values due to parallel database activity.

The design of this script was made to be non-agressive to the server performances.

@guettli
Copy link
Author

guettli commented Mar 2, 2019

Why not let the user decide?

You could provide a autodetect-ssd flag and a server-has-ssd flag.

Then the user can decide what he wants. There are arguments for both ways.

Then everyone would be happy.

@cherio
Copy link

cherio commented Apr 10, 2019

When scanning for "rotational" flag you filter out special directories ("." and "..") and "^sr" devices. It makes sense to add a filter for "^loop" devices as well. Those are pseudo devices.

@guettli
Copy link
Author

guettli commented Apr 11, 2019

Devices can have different names. Is it possible to measure the performance of the drive? This is the value which is all about. After measuring the performance the matching config value can get calculated.

@realulli
Copy link

Where do you draw the line? Imagine your VM running on a storage with spinning rust and half a TB of cache - are you planning to write a ton of data so a really large cache fills up?

Really, detection of SSDs is unreliable when it's hidden behind a virtualization layer.

@guettli
Copy link
Author

guettli commented Oct 30, 2019

@realulli Why not check the performance of the disk and set the values according the this value (float value, not boolean)?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants