When our application run for some time, for example , run for hours, the sbcl will throw heap exhausted exception.
Heap exhausted during garbage collection: 1968 bytes available, 2128 requested.
Gen StaPg UbSta LaSta LUbSt Boxed Unboxed LB LUB !move Alloc Waste Trig WP GCs Mem-age
0: 0 0 0 0 0 0 0 0 0 0 0 5368709 0 0 0.0000
1: 0 0 0 0 0 0 0 0 0 0 0 5368709 0 0 0.0000
2: 0 0 0 0 0 0 0 0 0 0 0 5368709 0 0 0.0000
3: 101912 101913 0 0 19362 20536 0 0 0 162867456 554752 102714709 0 1 1.4405
4: 130984 131071 0 0 29240 18868 0 0 25 191196152 5854216 128537781 14785 1 0.6442
5: 75511 81013 0 0 16567 17127 92 99 36 132974568 5818392 2000000 16565 0 0.0000
6: 0 0 0 0 7949 1232 0 0 0 37605376 0 2000000 7766 0 0.0000
Total bytes allocated = 524643552
Dynamic-space-size bytes = 536870912
GC control variables:
*GC-INHIBIT* = true
*GC-PENDING* = true
*STOP-FOR-GC-PENDING* = false
fatal error encountered in SBCL pid 3281(tid 3067845440):
Heap exhausted, game over.
Welcome to LDB, a low-level debugger for the Lisp runtime environment.
ldb>
Any suggestion?
SBCL does not allow you to allocate more than (sb-ext:dynamic-space-size) bytes on the heap. Here you have a 512MB default size (536870912 bytes) and the Lisp program already was using nearly that amount when it attempted to make another allocation.
You could double the amount of heap space available to 1024MB by starting SBCL with --dynamic-space-size 1024. However, as several comments point out, there may be a memory leak, where objects are referenced somehow proportional to the time that the system has been running, so this will offer only a temporary respite.
The (room t) standard Common Lisp function call might help debug this, if you call it periodically.
More advanced code like this http://dwim.hu/darcsweb/darcsweb.cgi?r=HEAD%20hu.dwim.debug;a=headblob;f=/source/path-to-root.lisp#l42 which delves into the SB-VM internal map of allocations could shed more light, and SBCL has a statistical profiler, http://www.sbcl.org/manual/#Statistical-Profiler that supports reporting on allocations too.
Related
I am trying to solve the linear system of equations A'*x = B using Matlab's "mldivide" (the backslash operator) in the form:
x_transp = A'\b;
A is a square sparse matrix and that is all I know about it.
The problem is that the transpose has no effect at all, so the result of the previous line of code is the same than:
x = A\b;
So, x = x_transp. However, either if I use a new variable such that:
A_transp = A';
x_transpOK1 = A_transp\b;
or simply use:
x_transpOK2 = transp(A)\b;
the result is different (x_transpOK1 = x_transpOK2 ≠ x = x_trans).
This behavior occurs in Matlab version 7.9.0 (R2009b) but it does not happen in 7.12 (R2011a).
This, however, does not happen with silly examples I have tried (the behavior then is correct). The matrices that make this behavior arise are:
A =[0.01 -0.495 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0
0 -0.495 0 0 1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 1];
b = [8
4
0
0
0
0
0
0
0
0];
Is it some kind of precision issue? Am I making any fundamental error I cannot see?
The guys at Mathworks replied: it is a bug in the interpreter, which have been fixed in the next versions. There is no fix for 7.9.0 and they recommend the following workaround:
A_transp = A';
x = A_transp\b;
I guess this is a great example of the typical advice to always be up-to-date...
My original post on Matlab Answers
The bug report
After all the discussion, here is my answer:
#Mario_Exec.bat, it seems to me that you might want to take this to the Matlab Answers (mathworks.com/matlabcentral/answers) as maybe someone with knowledge of the actual code (ie a Matlab employee) might be able to help you more specifically. It is an interesting question but it seems there is more going on that might need more knowledge of the actual code and decision trees.
Please do post back here when you hear back. I am curious what they say!
I use MongoDB to store price events for stocks. Depending on what you want to screen, the number of event can rapidly grow to 1Go-2Go.
I run MongoDB on a single machine and it is taking longer and longer to load the data. I am not able to find a clear answer on the web if "sharding on a single server" is a benefit to read speed.
Is it the right path to increase the read speed?
insert query update delete getmore command flushes mapped vsize res faults locked db
0 1 395 0 1 395 0 63.9g 128g 2.53g 7484 prices:70.5%
0 0 14726 0 5 14728 0 63.9g 128g 2.48g 31555 prices:7.4%
0 0 0 0 0 1 0 63.9g 128g 2.48g 436 prices:0.0%
0 0 0 0 1 1 0 63.9g 128g 2.48g 0 prices:0.0%
0 0 0 0 0 1 1 63.9g 128g 2.49g 3877 .:83.9%
0 0 0 0 0 1 0 63.9g 128g 2.49g 0 prices:0.0%
In Perl, I generate a huge read-only data-structure once, then fork().
This is to take advantage of COW on RSS pages when forking. It works really well, but when a child process exits, it allocates all the RAM from itelf just prior dying.
Is there a way to avoid this useless allocation ?
Here is sample Perl code that shows the issue.
#! /usr/bin/perl
my $a = [];
# Allocate 100 MiB
for my $i (1 .. 100000) {
push #$a, "x" x 1024;
}
# Fork 10 other process
for my $j (1 .. 10) {
last unless fork();
}
# Sleep for a while to be able to see the RSS
sleep(5);
In the sample vmstat output, we can see that it first allocates only 100MiB, then after the 1rst sleep it allocates the whole for a short while, and then releases all of it.
procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu------
r b swpd free buff cache si so bi bo in cs us sy id wa st
0 0 0 1329660 80596 86936 0 0 21 18 160 25 0 0 100 0 0
1 0 0 1328048 80596 86936 0 0 0 0 1013 44 0 0 100 0 0
0 0 0 1223888 80596 86936 0 0 0 0 1028 76 11 5 84 0 0
0 0 0 1223888 80596 86936 0 0 0 0 1010 40 0 0 100 0 0
0 0 0 1223888 80596 86936 0 0 0 0 1026 54 0 0 100 0 0
0 0 0 1223888 80596 86936 0 0 0 0 1006 39 0 0 100 0 0
13 0 0 741156 80596 86936 0 0 0 0 1012 66 13 58 28 0 0
0 0 0 1329288 80596 86936 0 0 0 0 1032 60 0 0 100 0 0
Note: it seems it isn't a Perl version specific issue. As I tested 5.8.8, 5.10.1 & 5.14.2 and they all do exhibit this behavior.
Update:
As #choroba asked in comments, I also tried to undef the data-structure, but it seems that it triggers the memory-touching as the RAM is then allocated.
You can add the following snippet at the end of the first script.
# Unallocate $a
undef $a;
# Sleep for a while to be able to see the RSS
sleep(5);
Actually, as I found out myself, this behavior is a feature, and the answer lies in the Perl doc:
The exit() function does not always exit immediately.
Likewise any object destructors that need to be called
are called before the real exit.
If this is a problem, you can
call POSIX::_exit($status) to avoid END and destructor processing.
And indeed, adding it at the end of the original code sample does avoid the behavior.
# XXX - To be added just before ending the process
# Use POSIX::_exit($status) to end without allocating copy-on-write RAM
use POSIX;
POSIX::_exit(0);
Note: for this to work, the child has to exit also before the data-structure goes out of scope.
I have wrote a test script which did millions of updates(using update query) in a collection. Following is the mongostat output
insert query update delete getmore command flushes mapped vsize res faults locked % idx miss % qr|qw ar|aw netIn netOut conn time
0 0 21156 0 0 1 0 208m 2.45g 119m 0 81.7 0 0|8 0|9 2m 1k 10 12:52:11
0 0 20620 0 0 1 0 208m 2.45g 119m 0 82.5 0 0|8 0|9 1m 1k 10 12:52:12
0 0 21915 0 0 1 0 208m 2.45g 119m 0 81.9 0 0|8 0|9 2m 1k 10 12:52:13
0 0 21634 0 0 1 0 208m 2.45g 119m 0 82.1 0 0|8 0|9 2m 1k 10 12:52:15
0 0 19793 0 0 1 0 208m 2.45g 119m 0 81.8 0 0|8 0|9 1m 1k 10 12:52:16
0 0 22062 0 0 1 0 208m 2.45g 119m 0 81.9 0 0|8 0|8 2m 1k 10 12:52:17
0 0 23395 0 0 1 0 208m 2.45g 119m 0 81.9 0 0|8 0|8 2m 1k 10 12:52:19
The netIn says the total network in bytes per second, i hope. Is there any way to increase the size of netIn to some mb, so that i can increase the update statement per second.
I don't think you understand the netIn statistic. It isn't some limit but the actual amount of data received by MongoDB per interval sample. In other words, the netIn value will increase if you (can) do more updates.
Increasing update throughput itself may be possible but is very application specific.
i have an installation on memcache which i want to use in my production environment but when i have ran a couple of tests it seems that memcache doesn't free up memory even after it has used up all of it allocated memory, Also i logged in and ran a flush_all command but the objects are still in the cache.
Here are outputs from some tests
memcached-tool
memcache-top v0.6 (default port: 11211, color: on, refresh: 3 seconds)
INSTANCE USAGE HIT % CONN TIME EVICT/s READ/s WRITE/s
127.0.0.1:11211 427.1% 0.0% 18 1.4ms 0.0 244 261.0K
AVERAGE: 427.1% 0.0% 18 1.4ms 0.0 244 261.0K
TOTAL: 4.3MB/ 1.0MB 18 1.4ms 0.0 244 261.0K
memcached-tool 127.0.0.1:11211 display
No Item_Size Max_age Pages Count Full? Evicted Evict_Time OOM
1 560B 4s 1 1872 yes 0 0 15488
2 704B 32s 1 559 no 0 0 0
3 880B 4s 1 1191 yes 0 0 1335
4 1.1K 9s 1 116 no 0 0 0
5 1.4K 21s 1 14 no 0 0 0
6 1.7K 4s 1 17 no 0 0 0
7 2.1K 84s 1 24 no 0 0 0
8 2.7K 130s 1 60 no 0 0 0
9 3.3K 25s 1 290 no 0 0 0
10 4.2K 9s 1 194 no 0 0 0
11 5.2K 9s 1 116 no 0 0 0
15 12.7K 816s 1 1 no 0 0 0
16 15.9K 769s 1 5 no 0 0 0
18 24.8K 786s 1 1 no 0 0 0
21 48.5K 816s 1 1 no 0 0 0
memcached-tool 127.0.0.1:11211 stats
127.0.0.1:11211 Field Value
accepting_conns 1
auth_cmds 0
auth_errors 0
bytes 4478060
bytes_read 23964596
bytes_written 546642860
cas_badval 0
cas_hits 0
cas_misses 0
cmd_flush 0
cmd_get 240894
cmd_set 4504
conn_yields 0
connection_structures 21
curr_connections 18
curr_items 4461
decr_hits 0
decr_misses 0
delete_hits 0
delete_misses 0
evictions 0
get_hits 43756
get_misses 197138
incr_hits 0
incr_misses 0
limit_maxbytes 1048576
listen_disabled_num 0
pid 8731
pointer_size 64
reclaimed 0
rusage_system 5.047232
rusage_user 4.311344
threads 4
time 1306247929
total_connections 3092
total_items 4504
uptime 1240
version 1.4.5
-m tells memcached how much RAM to use for item storage (in megabytes). Note
carefully that this isn't a global
memory limit, so memcached will use a
few % more memory than you tell it to.
Set this to safe values. Setting it to
less than 48 megabytes does not work
properly in 1.4.x and earlier. It will
still use the memory.
Source: https://github.com/memcached/memcached/wiki/ConfiguringServer#commandline-arguments