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

install/upgrade/outdated - improve performance and optimize queries #1397

Closed
wickles opened this issue Aug 31, 2017 · 20 comments
Closed

install/upgrade/outdated - improve performance and optimize queries #1397

wickles opened this issue Aug 31, 2017 · 20 comments

Comments

@wickles
Copy link

wickles commented Aug 31, 2017

Don't know if this is expected behavior or it's just my system, but choco outdated always takes a good 45 seconds to run fully, with 136 packages installed. It seems like it's querying for new data for each package, every time. Most package managers have a mypacman update style command to update a local cache of complete package data all at once, so that we don't need to query every package every time, and this command itself is generally much quicker than choco outdated for other package managers. It would be great if we could get a similar feature or some other way to speed up choco outdated.


NOTE: Added by admin.

Seeing Errors in New Versions of Choco?

If you find this causes issues with your repository of choice, please see #1770

Known Affected Servers include LaGet and NuGet Simple Server (PHP). Please see #1770 for those items as that is where that will be updated.

Release Notes

So we've implemented some optimizations for install/upgrade/outdated on getting packages. It is largely seen in outdated, where it is an order of magnitude faster (but it also has the most queries).

For Chocolatey v0.10.12+, we have converted to a much more efficient way of finding packages that allows us to reduce bandwidth and really speed things up. To do so, it means a change from FindPackagesById() to Packages() with a filter when we have a DataServicePackageRepository (aka a NuGet OData Package Repository Atom Feed).

The prior method FindPackagesById() would return ALL versions of a package and then locally choco/NuGet would filter them. This as you know can cause multiple queries if you have a lot of versions of a package, and some really inefficient use of resources and time. Moving to Packages() filtering by package Id and the latest version available means 1 result is returned. It's quite efficient and quite a bit faster than the older method by an order of magnitude.

Outdated Measurements

choco outdated makes heavy use of calling for upgraded versions. I have a system with about 34 packages installed.

  • Chocolatey 0.10.11 - Measure-Command { choco outdated -r --ignore-unfound }

    TotalSeconds       : 5.4484086
    TotalMilliseconds : 5448.4086
    
    • Roughly 160 milliseconds per package
  • Chocolatey 0.10.13 - Measure-Command { choco outdated -r --ignore-unfound }

    TotalSeconds       : 2.6337265
    TotalMilliseconds : 2633.7265 
    
    • Roughly 77 milliseconds per package

That's over 100% faster if I can math correctly.

@ferventcoder
Copy link
Member

@wickles yes, it's on our list to implement package indexes. That is why when you run choco update you will get a deprecated message that mentions that the command is being deprecated and will come back for that.

@ferventcoder
Copy link
Member

If you think it's bad now, you never had to work with the old Posh version of Choco - it was more like 40 minutes. Really, really bad. We've been talking about adding package indexes since 2014. https://www.slideshare.net/ferventcoder/chocolatey-puppet-conf2014/27

@dionysius
Copy link

Am I allowed to kindly be vocal about this slowiness here? It bugs me quite some time every time I use choco, any command it is soooo slow :(

C:\Users\Dionysius
λ ptime choco upgrade chocolatey

ptime 1.0 for Win32, Freeware - http://www.pc-tools.net/
Copyright(C) 2002, Jem Berkes <[email protected]>

=== choco upgrade chocolatey ===
Chocolatey v0.10.8
Upgrading the following packages:
chocolatey
By upgrading you accept licenses for the packages.
chocolatey v0.10.8 is the latest version available based on your source(s).

Chocolatey upgraded 0/1 packages.
 See the log for details (C:\ProgramData\chocolatey\logs\chocolatey.log).

Enjoy using Chocolatey? Explore more amazing features to take your
experience to the next level at
 https://chocolatey.org/compare

Execution time: 86.252 s

@ferventcoder
Copy link
Member

@dionysius head to the logs folder and take a look at the timing of the output of each log message. Even better, post it to a gist and share that here. That will help us see what may be causing things to be slow for you. That is a very long time to be waiting on a command and definitely not normal. It's not quite the same as this issue, as choco outdated is network dependent and needs to make quite a few calls, but most commands should be done quite quickly.

@ferventcoder
Copy link
Member

The chocolatey.log file by the way, not the summary.

@dionysius
Copy link

dionysius commented Oct 22, 2017

Hi @ferventcoder, didn't expected that quick response. While my issue is a general issue, I'll better raise a separate issue for that.

But maybe I can help a little for the time beeing, I can contribute to this exact named issue I've encounter also for choco outdated:

λ ptime choco outdated

ptime 1.0 for Win32, Freeware - http://www.pc-tools.net/
Copyright(C) 2002, Jem Berkes <[email protected]>

=== choco outdated ===
Chocolatey v0.10.8
Outdated Packages
 Output is package name | current version | available version | pinned?

npcap|0.83|0.83|false

Chocolatey has determined 0 package(s) are outdated.
 1 package(s) had warnings.
Warnings:
 - npcap

Execution time: 107.479 s

(I'm aware why npcap isn't listed anymore, there's a thread for it elsewhere)

The gist of chocolatey.log: https://gist.github.com/dionysius/c973f254355b04d349e6c9124993e89a

And in theory my computer and network is not potato :)

@bcurran3
Copy link

It takes 45-48 seconds for choco outdated to give me results as well.

I've noticed a considerable slowdown in choco processing after the version header for some time. I believe it goes back to when the manifest issue was changed back to not requiring elevated rights.

@dionysius
Copy link

dionysius commented Oct 22, 2017

Oh, good to know. All my commands are in elevated cmd already, otherwise the "auto confirmation" would not apply. (I think it's the option called "allowGlobalConfirmation", but It was long time ago I activated that)

Edit: And maybe it is how it is then - if so, I just like to let you know that there are wishes for improvement :)

@bcurran3
Copy link

I retract my statement about all choco command being slow. There was a point that it seemed that way, but I think it went away with 10.7 or 10.8.

@dionysius - thanks for the clue on ptime. I like it. I used to use the timer function in 4DOS a lot and needed a replacement since I haven't used 4DOS in years.

just for comparison

ptime 1.0 for Win32, Freeware - http://www.pc-tools.net/
Copyright(C) 2002, Jem Berkes [email protected]

=== choco outdated ===
Chocolatey v0.10.8 Professional
Outdated Packages
Output is package name | current version | available version | pinned?

irfanviewshellextension|1.04|1.04|false
nirlauncher|1.20.16|1.20.17|false
realtek-audio-drivers|2.82|2.82|false

Chocolatey has determined 1 package(s) are outdated.
2 package(s) had warnings.
Warnings:

  • irfanviewshellextension
  • realtek-audio-drivers

Execution time: 78.027 s

@Ytrog
Copy link

Ytrog commented Dec 7, 2017

I only have 34 packages and choco outdated is also very slow for me. It used to be fast however. Nowadays I sometimes even have a timeout :(

@ferventcoder
Copy link
Member

Choco outdated is always going to be dependent on the network, at least until pkg indexes are implemented and even then in some way when the indexes update.

DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 19, 2018
Added get_outdated to NugetService for simplification.
"packageManager.SourceRepository.FindPackage" downloaded data about all versions of package instead of only the latest. Now "packageManager.SourceRepository.GetPackages()" is used that gets only the latest version of a package.
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 19, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 20, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 20, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 20, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 20, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 20, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 20, 2018
Fixed naming IsRepositoryRemote to repository_is_remote.
Call .ToList() on GetPackages() to prevent warning from nugget.
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 21, 2018
chocolatey#1398 IgnoreUnfoundPackagesOnUpgradeOutdated
The .ToLower() is not needed.
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 21, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 21, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 21, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 21, 2018
DamianMaslanka5 added a commit to DamianMaslanka5/choco that referenced this issue Feb 22, 2018
Added get_outdated to NugetService for simplifiation.
"packageManager.SourceRepository.FindPackage" downloaded data about all versions of a package instead of only the latest.
Now "packageManager.SourceRepository.GetPackages()" is used to get only the latest version of a package.
ferventcoder added a commit that referenced this issue Oct 12, 2018
Ensure the config is reset to a known point with each package being run
in an operation. Previously it was thought this was happening, but the
way the reset occurred was actually changing the original configuration
and the current. Ensure that is getting reset accordingly by doing the
deep copy to set config each time through.
ferventcoder added a commit that referenced this issue Oct 12, 2018
Search through each repo based on features, instead of looking at only
what all in aggregate have available. This allows searching for
prereleases when available even if not every repo supports them.
ferventcoder added a commit to ferventcoder/choco that referenced this issue Oct 12, 2018
* stable:
  (chocolateyGH-1397) Search each repo based on features
  (chocolateyGH-1397) Fix: Ensure config reset
  (chocolateyGH-1656) Ensure check is not case sensitive
@ferventcoder ferventcoder self-assigned this Oct 13, 2018
@ferventcoder
Copy link
Member

This has been completed to speed up queries - this is available now in the beta (choco upgrade chocolatey --pre) and will be in 0.10.12.

@dionysius
Copy link

dionysius commented Oct 14, 2018

Can see the improvement :)

λ ptime choco outdated

ptime 1.0 for Win32, Freeware - http://www.pc-tools.net/
Copyright(C) 2002, Jem Berkes <[email protected]>

=== choco outdated ===
Chocolatey v0.10.12-beta-3-gb65f586
Outdated Packages
 Output is package name | current version | available version | pinned?

[...]

Chocolatey has determined 10 package(s) are outdated.

Execution time: 68.250 s

Compared to my earlier try (should be the same package list) it has at almost halved 👍 (107s down to 68s)

@wickles
Copy link
Author

wickles commented Oct 14, 2018

Nice. Thanks for the update!

@ferventcoder ferventcoder changed the title choco outdated is slow outdated - choco outdated is slow Mar 7, 2019
@ferventcoder ferventcoder changed the title outdated - choco outdated is slow outdated - improve performance of choco outdated Mar 13, 2019
ferventcoder added a commit that referenced this issue Mar 22, 2019
When attempting packaging operations like install/upgrade and a
repositories fails, log a warning and continue on. #612 implemented
some of the fixes for ignoring failing repositories, but apparently
that didn't apply when querying those repositories outside of the
package manager that was implemented with #1397.
@ferventcoder ferventcoder changed the title outdated - improve performance of choco outdated install/upgrade/outdated - improve performance and optimize queries Mar 22, 2019
ferventcoder added a commit to ferventcoder/choco that referenced this issue May 26, 2019
When searching for packages based on id, instead of using the older
methods, use the optimizations if they are turned on in the
configuration. This follows chocolateyGH-1397 to ensure that info uses the same
algorithm.
@asheroto
Copy link

asheroto commented Oct 21, 2021

I know this is an old issue but it raises some questions...

  • Why does Chocolatey query each package one by one? Why not concurrently?
  • Why does the entire package information have to transfer for commands such as choco outdated?
  • Why not have one query to the server that requests version numbers for all requested packages?

Currently choco outdated will issue GET requests one-by-one such as...

/api/v2/Packages()?$filter=(tolower(Id) eq 'audacity') and IsLatestVersion
/api/v2/Packages()?$filter=(tolower(Id) eq 'autohotkey') and IsLatestVersion
/api/v2/Packages()?$filter=(tolower(Id) eq 'ccleaner') and IsLatestVersion

Instead, consider one consolidated request such as...
/api/v2/Packages()?$filter=('audacity','autohotkey','ccleaner')

Example returned information:

{
  "audacity": "3.0.5",
  "autohotkey": "1.1.33.10",
  "ccleaner": "5.86.9258"
}

I realize this would take some additional work on the server side, but it would

  1. Save a lot of traffic
  2. Be much faster for users because only 1 GET request would take place

In addition, the amount of data returned by the API seems unnecessary for simple operations such as choco outdated.

Example query - 6 KB of data

Considering each request returns an average of 6 KB of data...

I have 193 packages... so it transfers 1.158 MB, which isn't a huge deal in terms of size, but having to process 193 GET requests of 6 KB each...takes a little bit, and for certain functions such as choco outdated, we don't need most of the data that is returned in each request.... we just need the version number.

Even if the team didn't want to implement a multi-package search in the API, an alternative solution would be:

  • Return only the package version when requested
  • Run the GET commands concurrently through async methods

Perhaps there is a backstory I am unaware of, or perhaps this is available in the business version (I only use it for personal use), but I happened to be Googling about this topic and came across this issue.

Thoughts?

Thank you guys for all the work you do on Chocolatey!

@dionysius
Copy link

Fell free to hide this comment as this won't contribute to the issue.

I was forced to switch to another package manager called scoop because of this and I'm very happy with it.

@chocolatey chocolatey deleted a comment from asheroto Oct 21, 2021
@pauby
Copy link
Member

pauby commented Oct 21, 2021

@asheroto This issue is closed. Can you please open a new issue.

@asheroto
Copy link

No prob! 🙂

@UsernameIsAlreadyTakenByMe

I'm not sure which issue is solved now, but the "choco upgrade all" command became really fast in "chocolatey v2.2.0" :)
...I just wanted to give some hope for the people using google and come here :)

Thank you :)

@pauby
Copy link
Member

pauby commented Jul 28, 2023

Performance improvements were made in 2.1.0 and this issue helps choco upgrade all for 2.2.0.

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

No branches or pull requests

9 participants