| Test race condition? (unit test?) [message #175583] |
Mon, 10 October 2011 08:16  |
Simon
Messages: 29
Registered: February 2011
Karma: 0
|
Junior Member |
|
|
Hi,
I have written a very simple file based lock/unlock system.
the way it works is as follows.
process A looks for File 'X', if file 'X' does not exists it tries to
creates an empty file, 'Xa', and locks it, it then creates file 'X' and
deletes file 'Xa'.
process B is also looking for File 'X', if file 'X' does not exists it
tries to create 'Xa', on failure it waits for a seconds and tries to
obtain the lock again. Once/if the lock is obtained, it looks for file
'X' again and either uses it or creates it.
process C looks for file 'X', if file 'X' exists it will use it.
I try to write as much unit tests as possible and, as this is an
important process, I would like to test the process above.
1- Can you think of any way I could test the above? As far as I know you
cannot start multiple threads in php, (even in a dev environment).
2- Is there a better way of getting a lock like I am doing above?
Many thanks
Simon
|
|
|
|
| Re: Test race condition? (unit test?) [message #175588 is a reply to message #175583] |
Mon, 10 October 2011 12:36  |
Jerry Stuckle
Messages: 2598
Registered: September 2010
Karma: 0
|
Senior Member |
|
|
On 10/10/2011 4:16 AM, Simon wrote:
> Hi,
>
> I have written a very simple file based lock/unlock system.
> the way it works is as follows.
>
> process A looks for File 'X', if file 'X' does not exists it tries to
> creates an empty file, 'Xa', and locks it, it then creates file 'X' and
> deletes file 'Xa'.
>
> process B is also looking for File 'X', if file 'X' does not exists it
> tries to create 'Xa', on failure it waits for a seconds and tries to
> obtain the lock again. Once/if the lock is obtained, it looks for file
> 'X' again and either uses it or creates it.
>
> process C looks for file 'X', if file 'X' exists it will use it.
>
> I try to write as much unit tests as possible and, as this is an
> important process, I would like to test the process above.
>
> 1- Can you think of any way I could test the above? As far as I know you
> cannot start multiple threads in php, (even in a dev environment).
>
> 2- Is there a better way of getting a lock like I am doing above?
>
> Many thanks
>
> Simon
Race conditions are always difficult to duplicate and troubleshoot when
they occur. And although with some third party tools you could process
php scripts in different threads, you probably don't want to (locking
between threads is different than locking between processes). The way
to typically troubleshoot such problems is to run multiple copies of the
script/program with waits for input at critical places. Run the script
concurrently from multiple windows to emulate the system.
The real question here is - what is the problem you're trying to
resolve. Without knowing that, it's difficult to recommend if there is
a better way.
--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
jstucklex(at)attglobal(dot)net
==================
|
|
|
|
| Re: Test race condition? (unit test?) [message #175589 is a reply to message #175588] |
Mon, 10 October 2011 13:48 |
Simon
Messages: 29
Registered: February 2011
Karma: 0
|
Junior Member |
|
|
>
> The real question here is - what is the problem you're trying to
> resolve. Without knowing that, it's difficult to recommend if there is a
> better way.
>
Thanks for the reply. I am not trying to solve a problem per say.
I am 'just' trying to prevent duplicate work.
When a request for a certain page first comes, it make a few calls to
the db and then caches the page.
Once the page is cached, only one call needs to be made.
But I sometime get hundreds of calls for a certain page, all at the same
time.
So it would make no sense for all of them to create the very same cache
page.
It makes more sense for the first request to cache the page wile others
wait a few milliseconds for it.
By locking the page, the first requests is instructing others that the
work is already been done and they should wait a few milliseconds.
Every so often, (for various reasons), the cache is marked as 'dirty'
and a new cache will then be recreated.
If one request 'locks' the page, and a cache is available, then the
other requests will continue to serve the dirty page.
I just want to prevent the same cache been created by many requests at
the same time.
I hope this makes sense.
Simon
|
|
|
|
| Re: Test race condition? (unit test?) [message #175590 is a reply to message #175589] |
Mon, 10 October 2011 14:43 |
The Natural Philosoph
Messages: 993
Registered: September 2010
Karma: 0
|
Senior Member |
|
|
Simon wrote:
>>
>> The real question here is - what is the problem you're trying to
>> resolve. Without knowing that, it's difficult to recommend if there is a
>> better way.
>>
>
> Thanks for the reply. I am not trying to solve a problem per say.
> I am 'just' trying to prevent duplicate work.
>
> When a request for a certain page first comes, it make a few calls to
> the db and then caches the page.
> Once the page is cached, only one call needs to be made.
>
> But I sometime get hundreds of calls for a certain page, all at the same
> time.
> So it would make no sense for all of them to create the very same cache
> page.
>
> It makes more sense for the first request to cache the page wile others
> wait a few milliseconds for it.
> By locking the page, the first requests is instructing others that the
> work is already been done and they should wait a few milliseconds.
>
> Every so often, (for various reasons), the cache is marked as 'dirty'
> and a new cache will then be recreated.
> If one request 'locks' the page, and a cache is available, then the
> other requests will continue to serve the dirty page.
>
> I just want to prevent the same cache been created by many requests at
> the same time.
>
well if you are using a database, it will already have cached the data
comprising the page.
Unless your manipulation of that page is excessively memory or CPU
intensive, there is little point in caching the result too.
And that is probably something a proxy server does better, anyway.
> I hope this makes sense.
>
> Simon
|
|
|
|
| Re: Test race condition? (unit test?) [message #175593 is a reply to message #175589] |
Mon, 10 October 2011 17:42 |
Jerry Stuckle
Messages: 2598
Registered: September 2010
Karma: 0
|
Senior Member |
|
|
On 10/10/2011 9:48 AM, Simon wrote:
>>
>> The real question here is - what is the problem you're trying to
>> resolve. Without knowing that, it's difficult to recommend if there is a
>> better way.
>>
>
> Thanks for the reply. I am not trying to solve a problem per say.
> I am 'just' trying to prevent duplicate work.
>
> When a request for a certain page first comes, it make a few calls to
> the db and then caches the page.
> Once the page is cached, only one call needs to be made.
>
> But I sometime get hundreds of calls for a certain page, all at the same
> time.
> So it would make no sense for all of them to create the very same cache
> page.
>
> It makes more sense for the first request to cache the page wile others
> wait a few milliseconds for it.
> By locking the page, the first requests is instructing others that the
> work is already been done and they should wait a few milliseconds.
>
> Every so often, (for various reasons), the cache is marked as 'dirty'
> and a new cache will then be recreated.
> If one request 'locks' the page, and a cache is available, then the
> other requests will continue to serve the dirty page.
>
> I just want to prevent the same cache been created by many requests at
> the same time.
>
> I hope this makes sense.
>
> Simon
Are you sure you need this? Databases will cache the results and future
db requests for the same information are generally quite fast.
Are you having performance problems? And if so, have you looked at
where your performance problems are? And if so, have you tried to
optimize your code?
Depending on what's going on, trying to cache pages can actually add
overhead to the server instead of relieving it.
--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
jstucklex(at)attglobal(dot)net
==================
|
|
|
|
Calendar
Search
Help
Members
Register
Login
Home
]