List:Falcon Storage Engine« Previous MessageNext Message »
From:Kevin Lewis Date:June 23 2009 3:24pm
Subject:Re: Proposal: extension to Log interface to get error message written
independently of falcon-debug-mask
View as plain text  
This looks OK to push to me.

Olav Sandstaa wrote:
> Hi,
> 
> The only feedback on my latest proposal/patch was from Vlad where he 
> suggested to change the name of the new debug mask from LogFatalError to 
> LogAlwaysWrite in order to make it more general and usable in other 
> contexts than fatal errors. I have made a new patch where I have changed 
> this. This patch is available here:
> 
>  http://lists.mysql.com/commits/76920
> 
> It would be great to get an OK to push this into Falcon as it would help 
> getting some more information written in some crashing tests that I am 
> working on.
> 
> Thanks,
> Olav
> 
> Olav Sandstaa wrote:
>> Vladislav Vaintroub wrote:
>>> Olav Sandstaa wrote:
>>>> Hi,
>>>>
>>>> I have now implement this functionality. The main added 
>>>> functionality to the Log class is:
>>>>
>>>> 1. A new debug mask called LogFatalError:
>>>>
>>>> This debug mask will be on by default for the Falcon log. It will not
>>>> be disabled if the user provides a debug mask at startup or changes 
>>>> the debug mask for a running server.
>>>>
>>>> 2. A new log method Log::fatal():
>>>>
>>>> Log messages written using this method will always be written to
>>>> Falcon/MySQL's log file.
>>>>
>>>> Note that compare to my original proposal this debug mask will only 
>>>> be on for the  original Falcon logger (StorageInterface::logger), 
>>>> not for the "MySQL logger" (StorageInterface::mysqlLogger). Due to 
>>>> this the default log mask is not completely maintained inside the 
>>>> Log class but has to be set in when the Falcon logger is added to 
>>>> listen to the log (in ha_falcon.cpp).
>>>>
>>>> The patch is available here:
>>>>
>>>>   http://lists.mysql.com/commits/68742
>>>>
>>>> It would be great if someone could have a look at the patch and 
>>>> either give feedback or an OK to push it.
>>>>
>>>> Thanks,
>>>> Olav
>>>
>>> Olav, I think it is ok basically, but I'd prefer to maybe have it 
>>> differently named. It could be situations where one wants to write to 
>>> the log, even if there is no fatal error, but on some very 
>>> interesting event. Maybe LogFatalError should be LogAlwaysWrite?
>>
>> Thanks for the suggestion, Vlad. I agree with you, there are also "non 
>> fatal" situations where Falcon should write messages to the Log:: file 
>> without depending on the falcon_debug_mask set by the user.
>>
>> The reason I have called it Log::fatal() and LogFatalError is that 
>> what caused me to propose this in the first place was that I had seen 
>> a few times where Falcon experienced situations where it could not 
>> continue - and exited without "telling anything about why".
>>
>> But I am not sure if I think you suggestion to rename LogFatalError is 
>> the only alternative. I see basically two other situations in addition 
>> to "fatal situations" where Falcon always should write something to 
>> the Log:: , serious error conditions where it is able to continue 
>> (e.g., disk full situtations?) and places where we should provide 
>> "info" (ie. during recovery or when we do the initials startup). So 
>> what do you and other think about the following suggestion (the first 
>> is the name of the "log flag" and the second is a corresponding 
>> convenience method):
>>
>>    1. LogFatalError   and Log::fatal()
>>    2. LogError  and Log::error()
>>    3. LogInfo and Log::info()
>>
>> where all three "debug masks" where by default on? (note that LogInfo 
>> is already defined and turning it on by default would increase the 
>> "verbosity" probably too much and some of the exiting LogInfo messages 
>> had to be changed to LogDebug).
>>
>> Vlad's suggestion would probably be implemented as:
>>
>>    1. LogAlwaysWrite and Log::logAlways()
>>
>> Opinions?
>>
>> Olav
>>
>>
> 
> 
Thread
Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskOlav Sandstaa17 Oct
  • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskJim Starkey17 Oct
    • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskAnn W. Harrison17 Oct
    • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskOlav Sandstaa10 Mar
      • Please review implementation of WL#4380 (Hooks for crashing Falcon andits recovery)Vladislav Vaintroub10 Mar
      • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskVladislav Vaintroub10 Mar
        • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskOlav Sandstaa15 Mar
          • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskOlav Sandstaa23 Jun
            • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskKevin Lewis23 Jun
            • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskVladislav Vaintroub23 Jun
      • Re: Proposal: extension to Log interface to get error message writtenindependently of falcon-debug-maskJim Starkey10 Mar