On Thu, 22 Aug 2013, John P. Rouillard wrote:
> Hi all:
>
> I am looking for ideas on how to solve this problem.
>
> In trying to troubleshoot a problem with an application, I want to
> include the prior 10 minutes of iostat info along with the report. I
> have the rules to detect the problem in the application but I am not
> sure how to capture/report the last 10 minutes of iostat info.
>
> When I was working with single line vmstat info, I could use this
> little action snippet:
>
> action = add $1_vmstat_data $0 ; \
> getsize %size $1_vmstat_data; \
> lcall %gt10 %size -> ( sub { $_[0] > 10 } ); \
> while %gt10 ( shift $1_vmstat_data %discard; \
> getsize %size $1_vmstat_data; \
> lcall %gt10 %size -> ( sub { $_[0] > 10 } ); \
> ) ;
>
> to trim the _vmstat_data down to just 10 lines. Since I am shifting
> off the old data and adding the new data I get the last 10 minutes.
>
> I would like to do the same with the iostat data, but it's multiline
> data and I don't know how many lines will be in the data (it depends
> on the command line values, how many disks are on the servers etc). I
> know each report will start with the "avg-cpu:" header line (but a
> different pid).
>
>
> iostat[26896]: avg-cpu: %user %nice %system %iowait %steal %idle
> iostat[26896]: 1.33 0.00 0.15 0.48 0.00 98.05
> iostat[26896]:
> iostat[26896]: Device: tps Blk_read/s Blk_wrtn/s Blk_read
> Blk_wrtn
> iostat[26896]: sdc 92.15 7116.66 2915.91 158637856297
> 64998797929
> iostat[26896]: sdb 5.88 1.75 1081.30 38910034
> 24103365344
> iostat[26896]: sdd 68.13 18690.11 2497.23 416622358522
> 55666016464
> iostat[26896]: sda 1.40 35.79 28.88 797836090
> 643783302
> iostat[26896]: dm-0 0.32 6.70 0.35 149414746
> 7819728
> iostat[26896]: dm-1 0.56 2.39 2.10 53243752
> 46759664
> iostat[26896]: dm-2 0.06 0.44 0.48 9874906
> 10752024
> iostat[26896]: dm-3 3.40 26.02 25.49 580007202
> 568158440
> iostat[26896]: dm-4 0.06 0.23 0.46 5237498
> 10293136
> iostat[26896]:
>
> I want just the most recent 10 of these to be stored. If I knew there
> was always 14 lines in the report, I could use the action above and
> just keep 140 lines but I don't know the number of lines.
>
> The ideas I have come up with are to break into perl and access the
> context either directly (in the associative array) or by "copy"ing the
> the context to an action variable and passing the variable into the
> perl function and manipulating the data by splitting on "avg-cpu:"
> keeping the last 10 and joining the results back together with
> avg-cpu" between them. I do lose a bit of info but...
>
> Another idea was to keep a rotating set of contexts as a circular
> buffer and just prepend all 10 contexts to the report. These will
> likely be out of order but at least the info is all there.
>
> But these are both lousy ideas. Does anybody have a better idea?
I would be looking to limit based on time instead of number of events.
Something
like 'report the last 10 minutes of iostat' rather than 'report the last 10
iostat sets of output'.
I'd have to go back through my saved mail or the list archives, but I seem to
remember a recent discussion on how to do a time-based limit like this.
David Lang
------------------------------------------------------------------------------
Introducing Performance Central, a new site from SourceForge and
AppDynamics. Performance Central is your source for news, insights,
analysis and resources for efficient Application Performance Management.
Visit us today!
http://pubads.g.doubleclick.net/gampad/clk?id=48897511&iu=/4140/ostg.clktrk
_______________________________________________
Simple-evcorr-users mailing list
Simple-evcorr-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/simple-evcorr-users
