You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@spark.apache.org by MUHAMMED CAN BAKIR <ca...@trendify.com.tr> on 2024/04/03 12:41:39 UTC

Spark Save as Parquet

Hi,

I am working on PySpark project that works on Azure VM. I am trying to trigger my PySpark code by using endpoint. I have a problem about saving my pyspark table as parquet to my Azure VM. I run same code on my local machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried same code in Azure VM, it tooks more than 6 hours. What should I do to detect the reason of this problem?
[cid:467759aa-d48d-4ae2-b572-04cb2d6915b4]

Re: Spark Save as Parquet

Posted by Femi Anthony <fe...@gmail.com>.
As I said, try it first with no configuration settings to see if that
helps, and only add configuration settings as needed.

Femi

On Thu, Apr 4, 2024 at 3:57 AM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
wrote:

> Data has 3 million row and 30 columns. Same data saving takes only under 2
> minutes on my local machine. I think problem might be about my
> configuration as you said. And also, RAM usage of my Azure VM never
> decreasing after running this code even it finishes. "I added spark.stop()"
> at the end of code already.
>
> ------------------------------
> *Gönderen:* Femi Anthony <fe...@gmail.com>
> *Gönderildi:* 4 Nisan 2024 Perşembe 10:41
> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
> *Konu:* Re: Spark Save as Parquet
>
> And how large is the dataset that gets written ? One thing I would suggest
> is to try saving with minimal config settings and see how it behaves, then
> add the configs one by one if you feel they're needed. I'm not sure about
> all the options esp those with a LEGACY value. I would even try without
> setting the executor memory first.
>
> Femi
>
> On Thu, Apr 4, 2024 at 3:30 AM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
> Actually, when I am running it on my local machines Jupyter Notebook, It
> saves it on my computer. Otherwise, when I running it on Azure VM, it saves
> it on my virtual machine folders.
>
> ------------------------------
> *Gönderen:* Femi Anthony <fe...@gmail.com>
> *Gönderildi:* 4 Nisan 2024 Perşembe 10:28
> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
> *Konu:* Re: Spark Save as Parquet
>
> So are you writing to the same location in both cases, only that you're
> running the code in a Jupyter notebook in one case and the other on an
> Azure VM ?
>
> Femi
>
> On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
>
> Hi,
>
> I am working on PySpark project that works on Azure VM. I am trying to
> trigger my PySpark code by using endpoint. I have a problem about saving my
> pyspark table as parquet to my Azure VM. I run same code on my local
> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
> same code in Azure VM, it tooks more than 6 hours. What should I do to
> detect the reason of this problem?
>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>


-- 
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre
minds." - Albert Einstein.

Ynt: Spark Save as Parquet

Posted by MUHAMMED CAN BAKIR <ca...@trendify.com.tr>.
Data has 3 million row and 30 columns. Same data saving takes only under 2 minutes on my local machine. I think problem might be about my configuration as you said. And also, RAM usage of my Azure VM never decreasing after running this code even it finishes. "I added spark.stop()" at the end of code already.

[cid:a787f412-4e87-4acf-b9f8-068b87042f38]
________________________________
Gönderen: Femi Anthony <fe...@gmail.com>
Gönderildi: 4 Nisan 2024 Perşembe 10:41
Kime: MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
Bilgi: user@spark.apache.org <us...@spark.apache.org>
Konu: Re: Spark Save as Parquet

And how large is the dataset that gets written ? One thing I would suggest is to try saving with minimal config settings and see how it behaves, then add the configs one by one if you feel they're needed. I'm not sure about all the options esp those with a LEGACY value. I would even try without setting the executor memory first.

Femi

On Thu, Apr 4, 2024 at 3:30 AM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>> wrote:
Actually, when I am running it on my local machines Jupyter Notebook, It saves it on my computer. Otherwise, when I running it on Azure VM, it saves it on my virtual machine folders.

[cid:ii_18ea80bb0b7d13c1f3c2]
________________________________
Gönderen: Femi Anthony <fe...@gmail.com>>
Gönderildi: 4 Nisan 2024 Perşembe 10:28
Kime: MUHAMMED CAN BAKIR <ca...@trendify.com.tr>>
Bilgi: user@spark.apache.org<ma...@spark.apache.org> <us...@spark.apache.org>>
Konu: Re: Spark Save as Parquet

So are you writing to the same location in both cases, only that you're running the code in a Jupyter notebook in one case and the other on an Azure VM ?

Femi

On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>> wrote:

Hi,

I am working on PySpark project that works on Azure VM. I am trying to trigger my PySpark code by using endpoint. I have a problem about saving my pyspark table as parquet to my Azure VM. I run same code on my local machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried same code in Azure VM, it tooks more than 6 hours. What should I do to detect the reason of this problem?
[cid:ii_18ea80bb0b7a99ae27a1]


--
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre minds." - Albert Einstein.


--
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre minds." - Albert Einstein.

Re: Spark Save as Parquet

Posted by Femi Anthony <fe...@gmail.com>.
And how large is the dataset that gets written ? One thing I would suggest
is to try saving with minimal config settings and see how it behaves, then
add the configs one by one if you feel they're needed. I'm not sure about
all the options esp those with a LEGACY value. I would even try without
setting the executor memory first.

Femi

On Thu, Apr 4, 2024 at 3:30 AM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
wrote:

> Actually, when I am running it on my local machines Jupyter Notebook, It
> saves it on my computer. Otherwise, when I running it on Azure VM, it saves
> it on my virtual machine folders.
>
> ------------------------------
> *Gönderen:* Femi Anthony <fe...@gmail.com>
> *Gönderildi:* 4 Nisan 2024 Perşembe 10:28
> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
> *Konu:* Re: Spark Save as Parquet
>
> So are you writing to the same location in both cases, only that you're
> running the code in a Jupyter notebook in one case and the other on an
> Azure VM ?
>
> Femi
>
> On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
>
> Hi,
>
> I am working on PySpark project that works on Azure VM. I am trying to
> trigger my PySpark code by using endpoint. I have a problem about saving my
> pyspark table as parquet to my Azure VM. I run same code on my local
> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
> same code in Azure VM, it tooks more than 6 hours. What should I do to
> detect the reason of this problem?
>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>


-- 
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre
minds." - Albert Einstein.

Ynt: Spark Save as Parquet

Posted by MUHAMMED CAN BAKIR <ca...@trendify.com.tr>.
Actually, when I am running it on my local machines Jupyter Notebook, It saves it on my computer. Otherwise, when I running it on Azure VM, it saves it on my virtual machine folders.

[cid:a211b047-9120-4e00-9c2a-0d44b34d1337]
________________________________
Gönderen: Femi Anthony <fe...@gmail.com>
Gönderildi: 4 Nisan 2024 Perşembe 10:28
Kime: MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
Bilgi: user@spark.apache.org <us...@spark.apache.org>
Konu: Re: Spark Save as Parquet

So are you writing to the same location in both cases, only that you're running the code in a Jupyter notebook in one case and the other on an Azure VM ?

Femi

On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>> wrote:

Hi,

I am working on PySpark project that works on Azure VM. I am trying to trigger my PySpark code by using endpoint. I have a problem about saving my pyspark table as parquet to my Azure VM. I run same code on my local machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried same code in Azure VM, it tooks more than 6 hours. What should I do to detect the reason of this problem?
[cid:ii_18ea80100faa99ae27a1]


--
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre minds." - Albert Einstein.

Re: Spark Save as Parquet

Posted by Femi Anthony <fe...@gmail.com>.
So are you writing to the same location in both cases, only that you're
running the code in a Jupyter notebook in one case and the other on an
Azure VM ?

Femi

On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
can.bakir@trendify.com.tr> wrote:

>
> Hi,
>
> I am working on PySpark project that works on Azure VM. I am trying to
> trigger my PySpark code by using endpoint. I have a problem about saving my
> pyspark table as parquet to my Azure VM. I run same code on my local
> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
> same code in Azure VM, it tooks more than 6 hours. What should I do to
> detect the reason of this problem?
>


-- 
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre
minds." - Albert Einstein.

Re: Spark Save as Parquet

Posted by Ananth Tirumanur <ak...@gmail.com>.
Are you running this on local mode or cluster mode? Is the VM located in a
different region compared with the data you are reading/writing? Does the
VM have sufficient compute resources and is the VM running other processes
at the same time?

it would be helpful if you provide the spark configuration and the spark
code you are using. Monitor the Azure VM's resource utilization (CPU, RAM,
disk I/O, network) during the execution. Access the Spark UI (typically
available at http://<driver-node-ip-address>:4040) to inspect the job
execution details, data skew, and potential bottlenecks.

--
Thanks, Ananth

On Thu, Apr 4, 2024 at 3:56 AM Femi Anthony <fe...@gmail.com> wrote:

> Can you specify the code you're using in both cases ?
>
> Femi
>
> On Wed, Apr 3, 2024 at 3:11 PM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
>> No, only process time is increasing too much. It saves as parquet in more
>> than 6 hours.
>>
>> iOS için Outlook <https://aka.ms/o0ukef> uygulamasını edinin
>> ------------------------------
>> *Gönderen:* Femi Anthony <fe...@gmail.com>
>> *Gönderildi:* Wednesday, April 3, 2024 9:52:49 PM
>> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
>> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
>> *Konu:* Re: Spark Save as Parquet
>>
>> Did you get an error message? If so, what was it ?
>>
>> Femi
>>
>> On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
>> can.bakir@trendify.com.tr> wrote:
>>
>>
>> Hi,
>>
>> I am working on PySpark project that works on Azure VM. I am trying to
>> trigger my PySpark code by using endpoint. I have a problem about saving my
>> pyspark table as parquet to my Azure VM. I run same code on my local
>> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
>> same code in Azure VM, it tooks more than 6 hours. What should I do to
>> detect the reason of this problem?
>>
>>
>>
>> --
>> http://dataphantik.com
>>
>> "Great spirits have always encountered violent opposition from mediocre
>> minds." - Albert Einstein.
>>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>

Re: Spark Save as Parquet

Posted by Mich Talebzadeh <mi...@gmail.com>.
Hi,

I  believe similar question was asked a few days ago

So what is the inherent issue with this affecting write performance?
path = "extracted_data/data.parquet"
extracted_data = self.spark.sql("""some sql operations""")
extracted_data.write.option("compression",
"snappy").format("parquet").mode("overwrite").save(path)

We need to consider some factors here:

1) If the dataset being written is large, having a suitable number of
executors for your Azure VM's resources is crucial. Too few executors can
lead to slower write performance as tasks get distributed among fewer
processing units.
Check the number of executors using the Spark UI (port 4040) and adjust if
needed based on your VM configuration.  Actually in your configuration I
did not see the configured number of executors.
2) you are using snappy compression. While Snappy compression is generally
efficient, it still adds overhead during the write process. If write
performance is critical and you have sufficient storage space, consider
using compression formats with faster write speeds, like Parquet's built-in
compression (e.g., gzip). Omit the .option("compression", "snappy") to use
Parquet's default.
3)  the performance of the underlying file system can influence write
speeds. For large datasets or scalability concerns, consider using Azure
Blob Storage or HDFS, as they are designed for distributed storage with
better performance for large-scale data operations.
4) ensure proper partitioning of your Parquet data. Well-partitioned data
can significantly improve write performance, especially for large datasets,
by allowing parallel writes to multiple files. Explore partitioning
strategies based on your data's structure and query patterns.

OS commands on vm

   1. top
   2. jps
   3. ps -aux
   4. free -m
   5. vmstat

HTH

Mich Talebzadeh,
Technologist | Solutions Architect | Data Engineer  | Generative AI
London
United Kingdom


   view my Linkedin profile
<https://www.linkedin.com/in/mich-talebzadeh-ph-d-5205b2/>


 https://en.everybodywiki.com/Mich_Talebzadeh



*Disclaimer:* The information provided is correct to the best of my
knowledge but of course cannot be guaranteed . It is essential to note
that, as with any advice, quote "one test result is worth one-thousand
expert opinions (Werner  <https://en.wikipedia.org/wiki/Wernher_von_Braun>Von
Braun <https://en.wikipedia.org/wiki/Wernher_von_Braun>)".


On Thu, 4 Apr 2024 at 17:15, MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
wrote:

> I am using same write code for both cases. I am creating pyspark table by
> using pyspark.sql and assign it to varible.
>
> path = "extracted_data/data.parquet"
> extracted_data = self.spark.sql("""some sql operations""")
> extracted_data.write.option("compression", "snappy").format("parquet").
> mode("overwrite").save(path)
>
> Extracted data has 3million rows and 30 columns. And, you can see my spark
> configuration below:
>
> *self*.spark = SparkSession.builder.appName(*self*.appName).master(*self*.
> master).
> config('spark.sql.execution.arrow.pyspark.enabled', True)
> .config('spark.sql.session.timeZone', 'UTC') .config('spark.driver.memory
> ','4G')
> .config('spark.ui.showConsoleProgress', True)
> .config('spark.sql.repl.eagerEval.enabled', True)
> .config("spark.executor.memory", "12g").config("spark.driver.maxResultSize
> ", "4096m")
> .config("spark.sql.parquet.outputTimestampType", "TIMESTAMP_MICROS")
> .config("spark.sql.parquet.datetimeRebaseModeInWrite", "LEGACY")
> .config("spark.sql.parquet.datetimeRebaseModeInRead", "LEGACY").
> getOrCreate()
> *self*.spark.conf.set("spark.sql.parquet.int96RebaseModeInWrite", "LEGACY
> ")
> *self*.spark.conf.set("spark.sql.parquet.int96RebaseModeInRead", "LEGACY")
>
>
> Saving this extracted data takes more than 6 hours in my Azure VM.
>
> ------------------------------
> *Gönderen:* Femi Anthony <fe...@gmail.com>
> *Gönderildi:* 4 Nisan 2024 Perşembe 09:40
> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
> *Konu:* Re: Spark Save as Parquet
>
> Can you specify the code you're using in both cases ?
>
> Femi
>
> On Wed, Apr 3, 2024 at 3:11 PM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
> No, only process time is increasing too much. It saves as parquet in more
> than 6 hours.
>
> iOS için Outlook <https://aka.ms/o0ukef> uygulamasını edinin
> ------------------------------
> *Gönderen:* Femi Anthony <fe...@gmail.com>
> *Gönderildi:* Wednesday, April 3, 2024 9:52:49 PM
> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
> *Konu:* Re: Spark Save as Parquet
>
> Did you get an error message? If so, what was it ?
>
> Femi
>
> On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
>
> Hi,
>
> I am working on PySpark project that works on Azure VM. I am trying to
> trigger my PySpark code by using endpoint. I have a problem about saving my
> pyspark table as parquet to my Azure VM. I run same code on my local
> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
> same code in Azure VM, it tooks more than 6 hours. What should I do to
> detect the reason of this problem?
>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>

Ynt: Spark Save as Parquet

Posted by MUHAMMED CAN BAKIR <ca...@trendify.com.tr>.
I am using same write code for both cases. I am creating pyspark table by using pyspark.sql and assign it to varible.

path = "extracted_data/data.parquet"
extracted_data = self.spark.sql("""some sql operations""")
extracted_data.write.option("compression", "snappy").format("parquet").mode("overwrite").save(path)

Extracted data has 3million rows and 30 columns. And, you can see my spark configuration below:

self.spark = SparkSession.builder.appName(self.appName).master(self.master).
config('spark.sql.execution.arrow.pyspark.enabled', True)
.config('spark.sql.session.timeZone', 'UTC') .config('spark.driver.memory','4G')
.config('spark.ui.showConsoleProgress', True)
.config('spark.sql.repl.eagerEval.enabled', True)
.config("spark.executor.memory", "12g").config("spark.driver.maxResultSize", "4096m")
.config("spark.sql.parquet.outputTimestampType", "TIMESTAMP_MICROS")
.config("spark.sql.parquet.datetimeRebaseModeInWrite", "LEGACY")
.config("spark.sql.parquet.datetimeRebaseModeInRead", "LEGACY").getOrCreate()
self.spark.conf.set("spark.sql.parquet.int96RebaseModeInWrite", "LEGACY")
self.spark.conf.set("spark.sql.parquet.int96RebaseModeInRead", "LEGACY")


Saving this extracted data takes more than 6 hours in my Azure VM.

[cid:d39b596e-f245-4cd6-b3b1-62671fc0adad]
________________________________
Gönderen: Femi Anthony <fe...@gmail.com>
Gönderildi: 4 Nisan 2024 Perşembe 09:40
Kime: MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
Bilgi: user@spark.apache.org <us...@spark.apache.org>
Konu: Re: Spark Save as Parquet

Can you specify the code you're using in both cases ?

Femi

On Wed, Apr 3, 2024 at 3:11 PM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>> wrote:
No, only process time is increasing too much. It saves as parquet in more than 6 hours.

iOS için Outlook<https://aka.ms/o0ukef> uygulamasını edinin
________________________________
Gönderen: Femi Anthony <fe...@gmail.com>>
Gönderildi: Wednesday, April 3, 2024 9:52:49 PM
Kime: MUHAMMED CAN BAKIR <ca...@trendify.com.tr>>
Bilgi: user@spark.apache.org<ma...@spark.apache.org> <us...@spark.apache.org>>
Konu: Re: Spark Save as Parquet

Did you get an error message? If so, what was it ?

Femi

On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>> wrote:

Hi,

I am working on PySpark project that works on Azure VM. I am trying to trigger my PySpark code by using endpoint. I have a problem about saving my pyspark table as parquet to my Azure VM. I run same code on my local machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried same code in Azure VM, it tooks more than 6 hours. What should I do to detect the reason of this problem?
[cid:ii_18ea7d61203a99ae27a1]


--
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre minds." - Albert Einstein.


--
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre minds." - Albert Einstein.

Re: Spark Save as Parquet

Posted by Femi Anthony <fe...@gmail.com>.
Can you specify the code you're using in both cases ?

Femi

On Wed, Apr 3, 2024 at 3:11 PM MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
wrote:

> No, only process time is increasing too much. It saves as parquet in more
> than 6 hours.
>
> iOS için Outlook <https://aka.ms/o0ukef> uygulamasını edinin
> ------------------------------
> *Gönderen:* Femi Anthony <fe...@gmail.com>
> *Gönderildi:* Wednesday, April 3, 2024 9:52:49 PM
> *Kime:* MUHAMMED CAN BAKIR <ca...@trendify.com.tr>
> *Bilgi:* user@spark.apache.org <us...@spark.apache.org>
> *Konu:* Re: Spark Save as Parquet
>
> Did you get an error message? If so, what was it ?
>
> Femi
>
> On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
> can.bakir@trendify.com.tr> wrote:
>
>
> Hi,
>
> I am working on PySpark project that works on Azure VM. I am trying to
> trigger my PySpark code by using endpoint. I have a problem about saving my
> pyspark table as parquet to my Azure VM. I run same code on my local
> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
> same code in Azure VM, it tooks more than 6 hours. What should I do to
> detect the reason of this problem?
>
>
>
> --
> http://dataphantik.com
>
> "Great spirits have always encountered violent opposition from mediocre
> minds." - Albert Einstein.
>


-- 
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre
minds." - Albert Einstein.

Re: Spark Save as Parquet

Posted by Femi Anthony <fe...@gmail.com>.
Did you get an error message? If so, what was it ?

Femi

On Wed, Apr 3, 2024 at 11:58 AM MUHAMMED CAN BAKIR <
can.bakir@trendify.com.tr> wrote:

>
> Hi,
>
> I am working on PySpark project that works on Azure VM. I am trying to
> trigger my PySpark code by using endpoint. I have a problem about saving my
> pyspark table as parquet to my Azure VM. I run same code on my local
> machine's Jupyter Notebook it tooks only 60 seconds otherwise when I  tried
> same code in Azure VM, it tooks more than 6 hours. What should I do to
> detect the reason of this problem?
>


-- 
http://dataphantik.com

"Great spirits have always encountered violent opposition from mediocre
minds." - Albert Einstein.