[STACK-2356] a10-octavia retry too long - Githubissues

a10networks / acos-client

ACOS API Client

Apache License 2.0

35 stars 61 forks source link

[STACK-2356] a10-octavia retry too long #347

Closed ytsai-a10 closed 3 years ago

ytsai-a10 commented 3 years ago

Description

If Bug Fix:

Severity Level High
Issue Description
1. acos-client may takes longer than 300 second for retry.
2. And after acos-client retry, a10-octavia still do another round of retry. And cause the retry takes very very long.

a10-octavia PR: https://github.com/a10networks/a10-octavia/pull/428

Jira Ticket

https://a10networks.atlassian.net/browse/STACK-2356

Technical Approach

change acos-client retry mechanism to use time.time() to calculate actual retry time.
in a10-octavia, when acos-client already do the retry then adjust retries correctly.

Config Changes

This is only required if the config has been updated in this PR

Required: Provide example config snippet within pre block


[a10_controller_worker]
amp_active_retries = 59
amp_active_wait_sec = 10
amp_vcs_retries = 59
amp_vcs_wait_sec = 10

Test Cases

delete instance when openstack running following tasks VThunderComputeConnectivityWait, ConfigureaVCSBackup, VCSSyncWait and GetMasterVThunder
measure the a10-octavia retry time. And the retry time should follow configuration option values.

Manual Testing

rpdb on target tasks
delete instance when rpdb stop the process, and then continue it.
trace log to see how long it take for retry