Home » Archived » Memory Analyzer (MAT) » Allegedly invalid HPROF file
Allegedly invalid HPROF file [message #3719] |
Thu, 28 August 2008 15:07 ![Go to next message Go to next message](theme/Solstice/images/down.png) |
Eclipse User![Friend of Eclipse Friend](/donate/web-api/friends_decorator.php?email=) |
|
|
|
Originally posted by: joerg.von.frantzius.artnology.com
This is a multi-part message in MIME format.
--------------000200070100010203090801
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
Hi,
when opening an HPROF file, I get this error:
> Invalid HPROF file: Expected to read another 1,342,205,952 bytes,
> but only 367,266,432 bytes are available.
> Invalid HPROF file: Expected to read another 1,342,205,952 bytes,
> but only 367,266,432 bytes are available.
When I open the same file with jhat, I get a lot of warnings like this:
> WARNING: Failed to resolve object id 0x480dca08 for field key
> (signature L)
but eventually it finishes parsing without errors, saying
Snapshot
resolved.
Started HTTP server on port
7000
Server is ready.
So I guess it's not as invalid as memory-analyzer suggests?
Regards,
J
|
|
|
Re: Allegedly invalid HPROF file [message #3890 is a reply to message #3719] |
Fri, 29 August 2008 10:39 ![Go to previous message Go to previous message](theme/Solstice/images/up.png) ![Go to next message Go to next message](theme/Solstice/images/down.png) |
Krum Tsvetkov![Friend of Eclipse Friend](/donate/web-api/friends_decorator.php?email=krum.tsvetkov%40sap.com) Messages: 165 Registered: July 2009 |
Senior Member |
|
|
Hi Jörg,
What you have observed with JHat also shows that there are some
inconsistencies with the heap dump.
In the Memory Analyzer we have added this length-validity check
intentionally, so that we can recognize such inconsisteny earlier and
fail. As we didn't have this check in the past what happened was that one
runs into unexpected errors and behaviour at a later point of time.
Sometimes the parsing didn't work out, but failed first 10 or 15 minutes
later, i.e. one had to wait a pretty long time to see that the dump cannot
be parsed. And sometimes there were other errors or strange results during
the analysis, which were caused by the missing objects, classes, etc...
If we figure out that such inconsistencies appear often, but still the
heap dumps are usable, then we can think of some "softer" error handling
and try to work with the dump despite of the problems. However, our
experience so far showed the opposite.
Anyway, if you are interested to try, I could build you a class with this
check removed. Then you can try what happens afterwards.
Best Regards,
Krum
|
|
|
Re: Allegedly invalid HPROF file [message #3896 is a reply to message #3890] |
Fri, 29 August 2008 13:56 ![Go to previous message Go to previous message](theme/Solstice/images/up.png) ![Go to next message Go to next message](theme/Solstice/images/down.png) |
Eclipse User![Friend of Eclipse Friend](/donate/web-api/friends_decorator.php?email=) |
|
|
|
Originally posted by: joerg.von.frantzius.artnology.com
Hi Krum,
in the meantime I fed the same HPROF dump into YourKit, and it happily
analyzed it.
Thanks for your answer anyway!
Jörg
Krum Tsvetkov wrote:
> Hi Jörg,
>
> What you have observed with JHat also shows that there are some
> inconsistencies with the heap dump. In the Memory Analyzer we have
> added this length-validity check intentionally, so that we can
> recognize such inconsisteny earlier and fail. As we didn't have this
> check in the past what happened was that one runs into unexpected
> errors and behaviour at a later point of time. Sometimes the parsing
> didn't work out, but failed first 10 or 15 minutes later, i.e. one had
> to wait a pretty long time to see that the dump cannot be parsed. And
> sometimes there were other errors or strange results during the
> analysis, which were caused by the missing objects, classes, etc...
>
> If we figure out that such inconsistencies appear often, but still the
> heap dumps are usable, then we can think of some "softer" error
> handling and try to work with the dump despite of the problems.
> However, our experience so far showed the opposite. Anyway, if you are
> interested to try, I could build you a class with this check removed.
> Then you can try what happens afterwards.
>
> Best Regards,
> Krum
>
>
|
|
|
Re: Allegedly invalid HPROF file [message #3951 is a reply to message #3896] |
Mon, 01 September 2008 10:27 ![Go to previous message Go to previous message](theme/Solstice/images/up.png) ![Go to next message Go to next message](theme/Solstice/images/down.png) |
Krum Tsvetkov![Friend of Eclipse Friend](/donate/web-api/friends_decorator.php?email=krum.tsvetkov%40sap.com) Messages: 165 Registered: July 2009 |
Senior Member |
|
|
Hi,
Nice to hear you managed to analyze the problem. However, I still can't
decide how to follow up on this.
One possibility is that we erroneously read the length and that the file
is fully ok. I would have a look again at the coding, but I think without
a concrete hprof to reproduce the this, it would be difficult to find it.
Anyway I will try. Do you think it will be possible to provide me somehow
access to this heap dump?
The other possibility is that indeed the file is inconsistent. In this
case (looking at the output you provided) about 1Gb of data is missing. If
this is true, I am still not convinced that the analysis witout this 1Gb
is correct. But as I suggested, if you, the users of the tool, find the
that this is the better behaviour, then we should rethink our
implementation.
As you were able to perform the analysis - do you think that the dump was
correct?
Regards,
Krum
|
|
|
Re: Allegedly invalid HPROF file [message #3957 is a reply to message #3951] |
Mon, 01 September 2008 11:51 ![Go to previous message Go to previous message](theme/Solstice/images/up.png) ![Go to next message Go to next message](theme/Solstice/images/down.png) |
Eclipse User![Friend of Eclipse Friend](/donate/web-api/friends_decorator.php?email=) |
|
|
|
Originally posted by: joerg.von.frantzius.artnology.com
Hi,
unfortunately, I cannot give you the HPROF file as it may contain
sensitive information. Since I was able to do the analysis using YourKit
profiler, from my point of view the dump was correct.
Regards,
Jörg
Krum Tsvetkov wrote:
> Hi,
>
> Nice to hear you managed to analyze the problem. However, I still
> can't decide how to follow up on this.
>
> One possibility is that we erroneously read the length and that the
> file is fully ok. I would have a look again at the coding, but I think
> without a concrete hprof to reproduce the this, it would be difficult
> to find it. Anyway I will try. Do you think it will be possible to
> provide me somehow access to this heap dump?
>
> The other possibility is that indeed the file is inconsistent. In this
> case (looking at the output you provided) about 1Gb of data is
> missing. If this is true, I am still not convinced that the analysis
> witout this 1Gb is correct. But as I suggested, if you, the users of
> the tool, find the that this is the better behaviour, then we should
> rethink our implementation.
>
> As you were able to perform the analysis - do you think that the dump
> was correct?
>
> Regards,
> Krum
>
|
|
| |
Goto Forum:
Current Time: Wed Aug 28 02:40:12 GMT 2024
Powered by FUDForum. Page generated in 0.05727 seconds
|