JEP 271: Unified GC Logging - Third pre-review

Kirk Pepperdine kirk.pepperdine at gmail.com
Mon Nov 16 12:44:04 UTC 2015


Hi Bengt,

I’m sure that I don’t care that 5 is missing nor that 3 comes after 5 and a few of the other collection phases are inter-weaved. I just wanted to make sure that something wasn’t missed. 

FYI, I’ve started parsing the log files… not that I am overly thrilled with the format (way too many lines IMO) but on the assumption this is as good as it’s going to get. I’d be super happy if we could get the header back.

Regards,
Kirk

> On Nov 16, 2015, at 1:17 PM, Bengt Rutisson <bengt.rutisson at oracle.com> wrote:
> 
> 
> Hi Kirk,
> 
> On 2015-11-13 19:49, Kirk Pepperdine wrote:
>> Hi Bengt,
>> 
>> GC(5) seems to be missing. from http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.02/logs/ConcMarkSweep-Xlog-gc-ergo-star-trace.txt <http://cr.openjdk.java.net/%7Ebrutisso/JEP-271/pre-review.02/logs/ConcMarkSweep-Xlog-gc-ergo-star-trace.txt>
> Good catch. This is a problem with the way CMS (or rather GenCollectedHeap) can skip a young GC and directly kick off a full collection if it finds that the young collection would fail. The way GenCollectedHeap::do_collection() handles the two types of collections that it can start is too complex in my opinion. 
> 
> This is the same code that makes this log:
> 
> http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.02/logs/Serial-Xlog-gc-info.txt <http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.02/logs/Serial-Xlog-gc-info.txt>
> 
> look like collection number 5 and 3 come in the wrong order.
> 
> (See the "Even more details..." section in my first pre-review: http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.00/compare.html <http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.00/compare.html>)
> 
> I will file an RFE to fix this but it is not really a logging problem. It is more of a code structure problem. So, I would like to fix this separately after the big unified logging change has been pushed. The new logging just makes it more visible.
> 
> Thanks,
> Bengt
> 
> 
>> 
>> Regards,
>> Kirk
>> 
>>> On Nov 13, 2015, at 9:56 AM, Bengt Rutisson <bengt.rutisson at oracle.com <mailto:bengt.rutisson at oracle.com>> wrote:
>>> 
>>> 
>>> Hi everyone,
>>> 
>>> Here is an updated pre-review for the work for JEP 271 Unified GC Logging.
>>> 
>>> Now it includes all the logging that I plan on converting for the first step.
>>> 
>>> Here are examples from the latest prototype:
>>> http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.02/compare.html <http://cr.openjdk.java.net/%7Ebrutisso/JEP-271/pre-review.02/compare.html>
>>> 
>>> The webrev for the changes is available here:
>>> http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.02/webrev.02/ <http://cr.openjdk.java.net/~brutisso/JEP-271/pre-review.02/webrev.02/>
>>> 
>>> Some changes that have been made:
>>> 
>>> - GC id is now logged as "GC(5)" instead of the previous "GC#5"
>>> - The CMS and G1 concurrent logging have been homogenized to look similar. All concurrent phases are start with "Concurrent" and now include timestamps.
>>> - The paused phases now all start with "Pause".
>>> - The log messages from the different collectors now look more similar. For example "Pause Young" is used for young GCs for all collectors instead of for example "Young pause" for G1 and "Pause. ParNew" for CMS.
>>> - I've added a line at the start of the log to clearly state what GC is selected.
>>> 
>>> Thanks,
>>> Bengt
>>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/attachments/20151116/dc705688/attachment.html>


More information about the hotspot-gc-dev mailing list