[Tarantool-patches] [PATCH v1] gitlab-ci: parallelize perf testing

[Tarantool-patches] [PATCH v1] gitlab-ci: parallelize perf testing

[Oleg Piskunov]

[-- Attachment #1: Type: text/plain, Size: 2410 bytes --]



- Adding 2 servers (sh1, sh2) for performance testing.
  Additional gitlab-ci tags for perf testing:
  docker_sh1_perf and docker_sh2_perf.
- Rebalance performance testing between servers.
- Changing gitlab-ci tag for performance docker images
  from 'perf' to 'deploy'.
 
Closes #4868
---
 
Github: https://github.com/tarantool/tarantool/tree/opiskunov/gh-4868-parallel-perf
Issue:  https://github.com/tarantool/tarantool/issues/4868
 
 .gitlab-ci.yml | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)
 
diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index ae2afce..897f657 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -91,8 +91,6 @@ variables:
   <<: *perf_only_definition
   image: ${IMAGE_PERF_BUILT}
   stage: perf
-  tags:
-    - docker_perf
   artifacts:
     when: always
     paths:
@@ -195,7 +193,7 @@ perf_bootstrap:
   <<: *perf_only_definition
   stage: test
   tags:
-    - perf
+    - deploy
   script:
     - ${GITLAB_MAKE} perf_prepare
 
@@ -203,18 +201,24 @@ perf_bootstrap:
 
 perf_sysbench:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'sysbench'
 
 perf_tpcc:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'tpcc'
 
 perf_ycsb_hash:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_sh2_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'ycsb'
@@ -222,6 +226,8 @@ perf_ycsb_hash:
 
 perf_ycsb_tree:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_sh2_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'ycsb'
@@ -229,6 +235,8 @@ perf_ycsb_tree:
 
 perf_nosqlbench_hash:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_sh1_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'nosqlbench'
@@ -236,6 +244,8 @@ perf_nosqlbench_hash:
 
 perf_nosqlbench_tree:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_sh1_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'nosqlbench'
@@ -243,6 +253,8 @@ perf_nosqlbench_tree:
 
 perf_cbench:
   <<: *perf_docker_test_definition
+  tags:
+    - docker_sh2_perf
   variables:
     <<: *perf_vars_definition
     BENCH: 'cbench'
--
1.8.3.1
 

[-- Attachment #2: Type: text/html, Size: 3692 bytes --]

Re: [Tarantool-patches] [PATCH v1] gitlab-ci: parallelize perf testing

[Alexander V. Tikhonov]

Hi Oleg, thanks for the patch. As I see you set the 2 new names for the
new tags, it ok. But you removed the default name from template and set
the rest tag with its default name, I think it's better to set this tag
in the new style, like docker_sh3_perf, otherwise better to avoid of
removement of the default value from template, and in this way no need
to set it manually to depend jobs. After this correction (choose any of
two described above) LGTM.

On Fri, May 08, 2020 at 11:58:04AM +0300, Oleg Piskunov wrote:
> 
> 
> - Adding 2 servers (sh1, sh2) for performance testing.
>   Additional gitlab-ci tags for perf testing:
>   docker_sh1_perf and docker_sh2_perf.
> - Rebalance performance testing between servers.
> - Changing gitlab-ci tag for performance docker images
>   from 'perf' to 'deploy'.
>  
> Closes #4868
> ---
>  
> Github: https://github.com/tarantool/tarantool/tree/opiskunov/gh-4868-parallel-perf
> Issue:  https://github.com/tarantool/tarantool/issues/4868
>  
>  .gitlab-ci.yml | 18 +++++++++++++++---
>  1 file changed, 15 insertions(+), 3 deletions(-)
>  
> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
> index ae2afce..897f657 100644
> --- a/.gitlab-ci.yml
> +++ b/.gitlab-ci.yml
> @@ -91,8 +91,6 @@ variables:
>    <<: *perf_only_definition
>    image: ${IMAGE_PERF_BUILT}
>    stage: perf
> -  tags:
> -    - docker_perf
>    artifacts:
>      when: always
>      paths:
> @@ -195,7 +193,7 @@ perf_bootstrap:
>    <<: *perf_only_definition
>    stage: test
>    tags:
> -    - perf
> +    - deploy
>    script:
>      - ${GITLAB_MAKE} perf_prepare
>  
> @@ -203,18 +201,24 @@ perf_bootstrap:
>  
>  perf_sysbench:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'sysbench'
>  
>  perf_tpcc:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'tpcc'
>  
>  perf_ycsb_hash:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_sh2_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'ycsb'
> @@ -222,6 +226,8 @@ perf_ycsb_hash:
>  
>  perf_ycsb_tree:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_sh2_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'ycsb'
> @@ -229,6 +235,8 @@ perf_ycsb_tree:
>  
>  perf_nosqlbench_hash:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_sh1_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'nosqlbench'
> @@ -236,6 +244,8 @@ perf_nosqlbench_hash:
>  
>  perf_nosqlbench_tree:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_sh1_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'nosqlbench'
> @@ -243,6 +253,8 @@ perf_nosqlbench_tree:
>  
>  perf_cbench:
>    <<: *perf_docker_test_definition
> +  tags:
> +    - docker_sh2_perf
>    variables:
>      <<: *perf_vars_definition
>      BENCH: 'cbench'
> --
> 1.8.3.1
>  

Re: [Tarantool-patches] [PATCH v1] gitlab-ci: parallelize perf testing

[Sergey Bronnikov]

On 11:58 Fri 08 May , Oleg Piskunov wrote:
> - Adding 2 servers (sh1, sh2) for performance testing.
>   Additional gitlab-ci tags for perf testing:
>   docker_sh1_perf and docker_sh2_perf.
> - Rebalance performance testing between servers.
> - Changing gitlab-ci tag for performance docker images
>   from 'perf' to 'deploy'.
>  
> Closes #4868

LGTM

Re: [Tarantool-patches] [PATCH v1] gitlab-ci: parallelize perf testing

[Oleg Piskunov]

[-- Attachment #1: Type: text/plain, Size: 3464 bytes --]


Sasha, thanks for review.
I will add specific tag for sh3 — «docker_sh3_perf»
  
>Среда, 13 мая 2020, 21:48 +03:00 от Alexander V. Tikhonov <avtikhon@tarantool.org>:
> 
>Hi Oleg, thanks for the patch. As I see you set the 2 new names for the
>new tags, it ok. But you removed the default name from template and set
>the rest tag with its default name, I think it's better to set this tag
>in the new style, like docker_sh3_perf, otherwise better to avoid of
>removement of the default value from template, and in this way no need
>to set it manually to depend jobs. After this correction (choose any of
>two described above) LGTM.
>
>On Fri, May 08, 2020 at 11:58:04AM +0300, Oleg Piskunov wrote:
>>
>>
>> - Adding 2 servers (sh1, sh2) for performance testing.
>>   Additional gitlab-ci tags for perf testing:
>>   docker_sh1_perf and docker_sh2_perf.
>> - Rebalance performance testing between servers.
>> - Changing gitlab-ci tag for performance docker images
>>   from 'perf' to 'deploy'.
>>  
>> Closes #4868
>> ---
>>  
>> Github:  https://github.com/tarantool/tarantool/tree/opiskunov/gh-4868-parallel-perf
>> Issue:  https://github.com/tarantool/tarantool/issues/4868
>>  
>>  .gitlab-ci.yml | 18 +++++++++++++++---
>>  1 file changed, 15 insertions(+), 3 deletions(-)
>>  
>> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
>> index ae2afce..897f657 100644
>> --- a/.gitlab-ci.yml
>> +++ b/.gitlab-ci.yml
>> @@ -91,8 +91,6 @@ variables:
>>    <<: *perf_only_definition
>>    image: ${IMAGE_PERF_BUILT}
>>    stage: perf
>> -  tags:
>> -    - docker_perf
>>    artifacts:
>>      when: always
>>      paths:
>> @@ -195,7 +193,7 @@ perf_bootstrap:
>>    <<: *perf_only_definition
>>    stage: test
>>    tags:
>> -    - perf
>> +    - deploy
>>    script:
>>      - ${GITLAB_MAKE} perf_prepare
>>  
>> @@ -203,18 +201,24 @@ perf_bootstrap:
>>  
>>  perf_sysbench:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'sysbench'
>>  
>>  perf_tpcc:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'tpcc'
>>  
>>  perf_ycsb_hash:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_sh2_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'ycsb'
>> @@ -222,6 +226,8 @@ perf_ycsb_hash:
>>  
>>  perf_ycsb_tree:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_sh2_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'ycsb'
>> @@ -229,6 +235,8 @@ perf_ycsb_tree:
>>  
>>  perf_nosqlbench_hash:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_sh1_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'nosqlbench'
>> @@ -236,6 +244,8 @@ perf_nosqlbench_hash:
>>  
>>  perf_nosqlbench_tree:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_sh1_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'nosqlbench'
>> @@ -243,6 +253,8 @@ perf_nosqlbench_tree:
>>  
>>  perf_cbench:
>>    <<: *perf_docker_test_definition
>> +  tags:
>> +    - docker_sh2_perf
>>    variables:
>>      <<: *perf_vars_definition
>>      BENCH: 'cbench'
>> --
>> 1.8.3.1
>>   
 
 
--
Oleg Piskunov
 

[-- Attachment #2: Type: text/html, Size: 5375 bytes --]