[ic] Upgrade to Interchange-5.7.4 performance issues

Stefan Hornburg (Racke) racke at linuxia.de
Wed Mar 17 14:26:28 UTC 2010


Bill Carr wrote:
> On Mar 17, 2010, at 9:53 AM, Stefan Hornburg (Racke) wrote:
> 
>> Bill Carr wrote:
>>> On Mar 17, 2010, at 7:43 AM, Stefan Hornburg (Racke) wrote:
>>>> Bill Carr wrote:
>>>>> On Mar 17, 2010, at 7:06 AM, Bill Carr wrote:
>>>>>> On Mar 16, 2010, at 11:14 PM, Peter wrote:
>>>>>>
>>>>>>> On 17/03/10 13:34, Bill Carr wrote:
>>>>>>>> I tried removing all UTF-8 flags and directives and I've tried with
>>>>>>>> 5.7.5 nightly and I still get big slow downs compared to running
>>>>>>>> 5.4.0. During the slow downs there are several interchange processes
>>>>>>>> consuming 100% CPU. What else can I try?
>>>>>>> Have you tried setting the MINIVEND_DISABLE_UTF8 environment variable
>>>>>>> before starting the Interchange server?
>>>>>> Yes, I put the following right in bin/interchange
>>>>>>
>>>>>> $ENV{MINIVEND_DISABLE_UTF8} = 1;
>>>>> I ran strace on both 5.4.0 and 5.7.5 and one obvious difference I see in 5.7.5 is the following:
>>>>> open("/dev/urandom", O_RDONLY)          = 18
>>>>> read(18, "\322\201\304#", 4)            = 4
>>>>> close(18)                               = 0
>>>>> select(8, [2], NULL, NULL, {1, 0})      = 1 (in [2], left {0, 999998})
>>>>> read(2, "a", 1)                         = 1
>>>>> select(8, [2], NULL, NULL, {1, 0})      = 1 (in [2], left {0, 999999})
>>>>> read(2, "r", 1)                         = 1
>>>>> select(8, [2], NULL, NULL, {1, 0})      = 1 (in [2], left {0, 999999})
>>>>> read(2, "g", 1)                         = 1
>>>>> select(8, [2], NULL, NULL, {1, 0})      = 1 (in [2], left {0, 999999})
>>>>> read(2, " ", 1)                         = 1
>>>>> select(8, [2], NULL, NULL, {1, 0})      = 1 (in [2], left {0, 999999})
>>>>> read(2, "0", 1)                         = 1
>>>>> Followed by hundreds of the last two lines above (with different values). Is that a clue? Is that expected?
>>>> It somehow reads some data through a socket one by one character, that's for certain not efficient.
>>>> Can you pastebin a complete strace so we can see which socket is attached to filehandle #8?
>>> I put here http://www.bottlenose-wine.com/userfiles/strace.txt
>>> It was too big for pastebin. Wasn't sure what the relevant parts were in order to pare it down.
>> What's your exact Interchange/Apache setup? Do you use by chance mod_interchange?
> Just like it says below:
> Apache/1.3.34 (Debian) mod_gzip/1.3.26.1a mod_ssl/2.8.25 OpenSSL/0.9.8c mod_interchange/1.33
> 
> I've never traced it but I also get poor performance running Apache2 with Interchange::Link

So it looks like your problem is related to mod_interchange / Interchange::Link.

Regards
	Racke



-- 
LinuXia Systems => http://www.linuxia.de/
Expert Interchange Consulting and System Administration
ICDEVGROUP => http://www.icdevgroup.org/
Interchange Development Team




More information about the interchange-users mailing list