[Fast/Warm restart] Implement helper class for waiting restart done #691
Conversation
Junchao-Mellanox
requested review from
arfeigin,
qiluo-msft,
stephenxs and
vaibhavhd
|
/azpw run Azure.sonic-swss-common |
|
/AzurePipelines run Azure.sonic-swss-common |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azpw run Azure.sonic-swss-common |
|
/AzurePipelines run Azure.sonic-swss-common |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Hi @qiluo-msft @liuh-80, could you please review and sign-off? |
…ther library to use
|
/azpw run Azure.sonic-swss-common |
|
/AzurePipelines run Azure.sonic-swss-common |
|
Azure Pipelines successfully started running 1 pipeline(s). |
common/restart_waiter.cpp
Outdated
| return false; | ||
| } | ||
|
|
||
| selectTimeout -= delay; |
There was a problem hiding this comment.
Not sure I understand. Let's say maxWaitTime is 90 seconds, thus initial selectTimeout is 90000 ms. For first iteration, restart is not done, and delay is 10000 ms. To make sure maxWaitTime is 90, the selectTimeout should be adjusted (90000 - 10000) = 80000 seconds because we have already waited 10 seconds.
|
/azpw run Azure.sonic-swss-common |
|
/AzurePipelines run Azure.sonic-swss-common |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azpw run Azure.sonic-swss-common |
|
/AzurePipelines run Azure.sonic-swss-common |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Hi @vaibhavhd @liuh-80 @stephenxs @arfeigin , could you please review and sign-off? |
Update sonic-swss-common submodule pointer to include the following: * abda263 Make the loglevel persistent by moving the LOGGER table from the LOGLEVEL DB to the CONFIG DB ([sonic-net#687](sonic-net/sonic-swss-common#687)) * d0fdf62 Check whether a pointer created by dynamic_cast is null before using it. ([sonic-net#689](sonic-net/sonic-swss-common#689)) * 2cae742 [Fast/Warm restart] Implement helper class for waiting restart done ([sonic-net#691](sonic-net/sonic-swss-common#691)) Signed-off-by: dprital <drorp@nvidia.com>
Depends on:
Why I did this?
Daemons which are not related to warm/fast restart might affect the performance of warm/fast restart. A hardcoded start up delay is the current solution to avoid this.
This PR implements a function to wait warm/fast restart done. This function provided a efficiency and graceful way for daemons to wait warm/fast restart done.
How I did it?
Implement a utility function RestartWaiter::waitRestartDone. This function waits warm restart done flag in STATE DB and return true if the flag is set by warm restart finalizer. This function is also exposed as python extension so that python daemons can utilize it.
This PR depends on new fastboot design: https://github.com/sonic-net/SONiC/blob/master/doc/fast-reboot/Fast-reboot_Flow_Improvements_HLD.md