﻿<?xml version="1.0" encoding="utf-8"?><rss version="2.0"><channel><title>Ayende @ Rahien</title><link>http://ayende.com</link><description>Ayende @ Rahien</description><copyright>Copyright (C) Ayende Rahien  2004 - 2021 (c) 2026</copyright><ttl>60</ttl><item><title>Sergey Shumov commented on Debugging memory issues with RavenDB using WinDBG</title><description>Ayende, have you tried to compare memory usage of ESENT vs Munin?</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment13</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment13</guid><pubDate>Fri, 07 Dec 2012 03:46:58 GMT</pubDate></item><item><title>Alois Kraus commented on Debugging memory issues with RavenDB using WinDBG</title><description>You should create several dumps while the memory is growing to see a pattern in one of the threads to see where it is actually allocating something. Only by looking at the threads while they are not allocating will not lead the way. With VMMap from SysInternals you can have a look at the contents and newly allocated stuff as well. That can also give some clue what the stuff taking so much space actually is. </description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment12</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment12</guid><pubDate>Thu, 06 Dec 2012 23:24:36 GMT</pubDate></item><item><title>Roman D. Boiko commented on Debugging memory issues with RavenDB using WinDBG</title><description>Note that using memprofiler with unmanaged resources tracker enabled to profile your process will give you more information than using it to analyze a dump file.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment11</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment11</guid><pubDate>Thu, 06 Dec 2012 19:58:16 GMT</pubDate></item><item><title>Roman D. Boiko commented on Debugging memory issues with RavenDB using WinDBG</title><description>Ayende, I had to mention that memprofiler partially analyses non-managed code. See http://memprofiler.com/OnlineDocs, "Native Memory Page" section. There is no guarantee that it will solve your particular issue, but it is free to try.

As for the book, it doesn't cover memprofiler in any detail, but provides several good strategies for memory profiling in general. Not sure they will be new for you, though.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment10</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment10</guid><pubDate>Thu, 06 Dec 2012 19:30:13 GMT</pubDate></item><item><title>James Manning commented on Debugging memory issues with RavenDB using WinDBG</title><description>A couple of tools I've found helpful fast/easy/simple to use for memory investigation:

- CLRProfiler4 (despite having a name that sounds more CPU-centric) from the CLR team (IIRC) - if it has any managed code, mainly because it's so simple and quick to use that it's useful even if only as a verification that the managed heap(s) are fine.  Its heap map is very useful, especially in comparison to windbg/sos :)

http://www.microsoft.com/en-us/download/details.aspx?id=16273

- Process Explorer - not as detailed as other tools, but its process properties dialog includes a simple way to see all memory info ('Performance' tab) and managed memory ('.NET Performance' tab, especially with '.NET CLR Memory', the default, selected under '.NET Performance Objects')

Much like the GitHub's situation with "the bridge loop problem that wasn't there" (https://github.com/blog/1346-network-problems-last-friday), I find it important to use the coarse-granularity tools available to help narrow down the problem (even if it's just to verify a 'hunch') before bringing out the tools like windbg :)

Of course, that's exactly what Ayende did here, of course, just reiterating that it's a great idea. :)</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment9</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment9</guid><pubDate>Thu, 06 Dec 2012 19:28:33 GMT</pubDate></item><item><title>Ayende Rahien commented on Debugging memory issues with RavenDB using WinDBG</title><description>Chris,
I am pretty sure it is an Esent issue, but I am also pretty sure that it isn't the cache. We added some diagnostics, and the cache takes ~400MB, so that isn't counting for a lot of the data, still investigating.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment8</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment8</guid><pubDate>Thu, 06 Dec 2012 19:01:55 GMT</pubDate></item><item><title>Ayende Rahien commented on Debugging memory issues with RavenDB using WinDBG</title><description>Roman,
I am pretty sure that the problem is not in managed code, so memprofiler isn't likely to help me here.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment7</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment7</guid><pubDate>Thu, 06 Dec 2012 18:59:33 GMT</pubDate></item><item><title>Matt commented on Debugging memory issues with RavenDB using WinDBG</title><description>My immediate thought when I saw this was Esent too.  Setting the cache parameters solved a similar problem for me.  I believe there are actually some config options for this in RavenDB, but you probably know that ;)</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment6</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment6</guid><pubDate>Thu, 06 Dec 2012 17:36:17 GMT</pubDate></item><item><title>Justin commented on Debugging memory issues with RavenDB using WinDBG</title><description>As your managed heap is very low, have you considered using DebugDiag's LeakTrack to try and catch native allocations that leak?</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment5</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment5</guid><pubDate>Thu, 06 Dec 2012 17:11:27 GMT</pubDate></item><item><title>Chris Eldredge commented on Debugging memory issues with RavenDB using WinDBG</title><description>Loading the dmp and using !address -summary shows 1.15 Gb of committed memory, but as you note the problem does not seem to be in a managed heap since !dumpheap, !eeheap and !heapstat all show memory consumption less than 100 Mb.

This seems to indicate most memory is being allocated and used by unmanaged code.

You mentioned that you are using ESENT. Did you see another user ran into a problem with Managed ESENT that caused over 1 GB of memory to be consumed? http://managedesent.codeplex.com/discussions/276175</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment4</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment4</guid><pubDate>Thu, 06 Dec 2012 15:50:44 GMT</pubDate></item><item><title>Ryan commented on Debugging memory issues with RavenDB using WinDBG</title><description>Keep in mind !dumpheap shows the managed heap.  It won't show you the unmanaged memory used by the process, and if you're using esent that points more fingers toward the issue being in unmanaged memory.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment3</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment3</guid><pubDate>Thu, 06 Dec 2012 15:03:25 GMT</pubDate></item><item><title>Rémi BOURGAREL commented on Debugging memory issues with RavenDB using WinDBG</title><description>windbg is not really user friendly. Even helped by  Tess's blog here http://blogs.msdn.com/b/tess/archive/tags/memory+issues/default.aspx?PageIndex=1 I can't figure it out.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment2</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment2</guid><pubDate>Thu, 06 Dec 2012 13:37:54 GMT</pubDate></item><item><title>Roman D. Boiko commented on Debugging memory issues with RavenDB using WinDBG</title><description>http://memprofiler.com is a great memory profiler. See also http://stackoverflow.com/questions/399847/net-memory-profiling-tools and http://www.apress.com/9781430244585 for alternatives.</description><link>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment1</link><guid>http://ayende.com/160193/debugging-memory-issues-with-ravendb-using-windbg#comment1</guid><pubDate>Thu, 06 Dec 2012 12:26:47 GMT</pubDate></item></channel></rss>