一:背景
1. 讲故事
上个月有位朋友wx找到我,说他的程序存在内存泄漏问题,寻求如何解决? 如下图所示:
从截图中可以看出,这位朋友对 windbg 的操作还是有些熟悉的,可能缺乏一定的实操经验,所以用了几个命令之后就不知道怎么排查下去了。
既然找到我,那就以我的个人经验在他的dump上继续分析寻找罪魁祸首,闲话不多说,上windbg说话。
二:Windbg 分析
1. 真的存在内存泄漏吗?
追这个系列的朋友应该知道,我无数次的用 !address -summary
和 !eeheap -gc
这两个命令来判断当前的内存泄漏是属于托管层还是非托管层。
0:000> !address -summary
--- Usage Summary ---------------- RgnCount ----------- Total Size -------- %ofBusy %ofTotal
Free 358 7dfc`67f60000 ( 125.986 TB) 98.43%
<unknown> 1087 203`88b6e000 ( 2.014 TB) 99.99% 1.57%
Image 1532 0`09f11000 ( 159.066 MB) 0.01% 0.00%
Heap 249 0`03453000 ( 52.324 MB) 0.00% 0.00%
Stack 66 0`01fc0000 ( 31.750 MB) 0.00% 0.00%
Other 10 0`001d1000 ( 1.816 MB) 0.00% 0.00%
TEB 22 0`0002c000 ( 176.000 kB) 0.00% 0.00%
PEB 1 0`00001000 ( 4.000 kB) 0.00% 0.00%
--- Type Summary (for busy) ------ RgnCount ----------- Total Size -------- %ofBusy %ofTotal
MEM_MAPPED 183 200`00d06000 ( 2.000 TB) 99.30% 1.56%
MEM_PRIVATE 1252 3`8d479000 ( 14.207 GB) 0.69% 0.01%
MEM_IMAGE 1532 0`09f11000 ( 159.066 MB) 0.01% 0.00%
--- State Summary ---------------- RgnCount ----------- Total Size -------- %ofBusy %ofTotal
MEM_FREE 358 7dfc`67f60000 ( 125.986 TB) 98.43%
MEM_RESERVE 749 200`28a9b000 ( 2.001 TB) 99.33% 1.56%
MEM_COMMIT 2218 3`6f5f5000 ( 13.740 GB) 0.67% 0.01%
0:000> !eeheap -gc
Number of GC Heaps: 1
generation 0 starts at 0x00000026DA8DA928
generation 1 starts at 0x00000026DA7FC348
generation 2 starts at 0x00000024C4691000
ephemeral segment allocation context: none
segment begin allocated size
00000024C4690000 00000024C4691000 00000024D468FF28 0xfffef28(268431144)
00000024EECF0000 00000024EECF1000 00000024FECF0000 0xffff000(268431360)
000000248D6F0000 000000248D6F1000 000000249D6EFEF8 0xfffeef8(268431096)
...
00000026D66D0000 00000026D66D1000 00000026DBA3CA30 0x536ba30(87472688)
Large object heap starts at 0x00000024D4691000
segment begin allocated size
00000024D4690000 00000024D4691000 00000024DC67C318 0x7feb318(134132504)
00000024E60F0000 00000024E60F1000 00000024EE0637C8 0x7f727c8(133638088)
0000002482140000 0000002482141000 000000248A08F338 0x7f4e338(133489464)
00000024A6770000 00000024A6771000 00000024AE76F6C0 0x7ffe6c0(134211264)
...
000000278E6D0000 000000278E6D1000 000000279635F2D0 0x7c8e2d0(130605776)
00000029233E0000 00000029233E1000 000000292AF672F8 0x7b862f8(129524472)
000000292B3E0000 000000292B3E1000 0000002931A5ED60 0x667dd60(107470176)
000000299B3E0000 000000299B3E1000 00000029A20095B0 0x6c285b0(113411504)
000000281E6D0000 000000281E6D1000 0000002825CD3F58 0x7602f58(123744088)
00000028266D0000 00000028266D1000 000000282D5CAD50 0x6ef9d50(116366672)
000000282E6D0000 000000282E6D1000 0000002833CA0880 0x55cf880(89979008)
00000029A33E0000 00000029A33E1000 00000029A684D300 0x346c300(54969088)
Total Size: Size: 0x353f96d88 (14293757320) bytes.
------------------------------
GC Heap Size: Size: 0x353f96d88 (14293757320) bytes.
从输出看,当前进程占用 MEM_COMMIT=13.7G
,托管堆内存占用 14293757320 = 13.3G
,很明显这属于简单模式的 托管内存泄漏
,根据经验,托管堆上可能有什么大对象,这里用 !dumpheap -stat
命令。
0:000> !dumpheap -stat
Statistics:
MT Count TotalSize Class Name
00007ff9ed6ea268 3956842 94964208 System.Collections.Generic.Dictionary`2+KeyCollection[[System.String, System.Private.CoreLib],[Serilog.Events.LogEventPropertyValue, Serilog]]
00007ff9ed5e6d28 3842435 166405016 Serilog.Parsing.MessageTemplateToken[]
00007ff9ed5e5e28 3842434 184436832 Serilog.Events.MessageTemplate
00007ff9ecccf090 4011012 203304420 System.Int32[]
00007ff9ed647078 3956849 253238336 Serilog.Events.LogEvent
00007ff9ed6e7b48 3956849 284893128 System.Collections.Generic.Dictionary`2[[System.String, System.Private.CoreLib],[Serilog.Events.LogEventPropertyValue, Serilog]]
00007ff9ed5e74e8 9259598 296307136 Serilog.Parsing.TextToken
00007ff9ed6471b0 12551808 301243392 Serilog.Events.ScalarValue
00007ff9ed6e8308 3956849 729078048 System.Collections.Generic.Dictionary`2+Entry[[System.String, System.Private.CoreLib],[Serilog.Events.LogEventPropertyValue, Serilog]][]
00007ff9eccb1e18 16546412 3987811940 System.String
00000024c3b8faf0 82904 7382993568 Free
我去,托管堆最大的对象居然是 Free
,大概占用 7.3G
, 这就