0

I am doing a Bitbucket Load test before enabling LFS in production.

Below are the Stats: We Simulated Production load on stage instance & I had selected a 30 MB repo to clone to capture time before performing LFS operations & after enabling LFS operations. This 30 MB repo doest not have any LFS files in it.

Clone detail Before LFS load of 3.4 GB

time git clone ssh://git@mycompayurl.com:7999/project1/loadtest_repo.git Cloning into 'loadtest_repo... remote: Enumerating objects: 26260, done. remote: Counting objects: 100% (26260/26260), done. remote: Compressing objects: 100% (9024/9024), done. remote: Total 26260 (delta 17450), reused 24565 (delta 16555)/s Receiving objects: 100% (26260/26260), 29.16 MiB | 188.00 KiB/s, done. Resolving deltas: 100% (17450/17450), done.

real 2m43.301s user 0m1.829s sys 0m1.306s

The clone time was 2m43 Sec ( on User Laptop) The server response time was 135957 (2.26 m)

172.19.241.253 | ssh | o*CVW0H6x1000x18849x15 | ravrao | 2020-04-02 16:42:23,265 | SSH - git-upload-pack '/project1/loadtest_repo.git' | "SSH-2.0-OpenSSH_7.9" | 0 | 4973 | 30623556 | cache:miss, clone, protocol:1, ssh:user:id:18160 | 135957 | 1j0141y |

After LFS Load.

time git clone ssh://git@staging.git.source.akamai.com:7999/project1/loadtest_repo.git Cloning into 'loadtest_repo'... remote: Enumerating objects: 26260, done. remote: Counting objects: 100% (26260/26260), done. remote: Compressing objects: 100% (9024/9024), done. remote: Total 26260 (delta 17450), reused 24565 (delta 16555)/s Receiving objects: 100% (26260/26260), 29.16 MiB | 123.00 KiB/s, done. Resolving deltas: 100% (17450/17450), done.

real 4m10.009s user 0m1.873s sys 0m1.296s

The clone time was 4m10 Sec (on User Laptop)

The server response time was 232.751 Sec 172.19.242.234 | ssh | o*CVW0H6x974x38452x115 | ravrao | 2020-04-03 16:18:30,283 | SSH - git-upload-pack '/project1/loadtest_repo.git' | "SSH-2.0-OpenSSH_7.9" | 0 | 4973 | 30617456 | cache:hit, clone, protocol:1, ssh:user:id:18160 | 232751 | c17fgf |

The server response time was increase to 71.85 % (135957 Mil Sec Before LFS - 232751 After LFS = 96794)

Question: 1. The server response time is more than the user response time.

I am trying to find the Root cause for increase in Server Response time.

Let us know what are the factors which cause git clone server response time to increase?

what might be best methods to do LFS load test.

Thanks in advance

rgh
  • 40
  • 2
  • 11

1 Answers1

0

After testing I found the Bitbucket application prints server response time in log only after below step completed.

Receiving objects: 100% (26260/26260), 29.16 MiB | 123.00 KiB/s, done.

Also I realised the server response time varies depending on the client network bandwidth.

Next question is, the tomcat / Bitbucket server response time is was beloved to be the time taken by the server to process the request however now it depends on network bandwidth too.

Is this applicable to HTTPs clone Operations too?

Can someone point me to server response time document.

Is server response time depends on client bandwidth (in my case for git operations)

rgh
  • 40
  • 2
  • 11