[systemd-devel] Hear opinions about changing watchdog timeout value during service running

[김민경/주임연구원/SW Platform(연)AOT팀(minkyung88.kim@lge.com)]

Hello.

Below is usecase from our SW-Engineer.

What do you think so?


 From Minkyung Kim.


The usecase:



When process runs, it will face various situation according to a job.

Process can send a notification fast when it’s workload is not heavy. On 
the contrary if the workload is heavy, sending notification will be delayed.

In the SW being developed currently, the gap between 2 cases are really big.

In the best case process can send a notification within 2 seconds, but 
as for the worst case 30 seconds is spent.

You may think like, SW should be modified to send notification as 
uniform as possible.

Unfortunately we can’t modify it because that part is placed in the 
library that we can’t change.



Or we can apply maximum time as a notification interval (In our case 30 
seconds).

But I think it could be better if we can change the timer value to come 
up with the specific time period when SW can send notification fast.

It could be more flexible because we can apply timer dynamically 
regarding the SW status.


2016-05-21 오전 1:59에 Lennart Poettering 이(가) 쓴 글:
> On Thu, 19.05.16 10:57, 김민경/주임연구원/SW Platform(연)AOT팀(minkyung88.kim at lge.com) (minkyung88.kim at lge.com) wrote:
>
>> Hello,
>>
>> I am planning to work on supporting api which changes "WATCHDOG_USEC"
>> value during service running.
>>
>> With this api, running service can change its own watchdog timeout value.
>>
>> I want to hear your opinions about this suggested function.
> Hmm, what's the usecase here?
>
> Adding this sounds simply enough: it could be as easy as introducing a
> WATCHDOG_USEC parameter for sd_notify(), which would then alter the
> watchdog settings...
>
> But before doing that, what's the background?
>
> Lennart
>