Okay, this is quite strange...
I was running a "vmstat 1" on a mail server here and typed "du -s *" in a
directory that a few directories, some with a whole bunch of files in
them. The "vmstat 1" showed:
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 0 0 2768 67348 176860 436628 0 0 1 0 261 268 9 8 84
1 0 0 2768 66876 176860 436628 0 0 0 0 217 176 14 8 78
1 0 0 2768 65760 176860 436628 0 0 0 0 199 174 53 6 40
3 0 0 2768 66492 176860 436628 0 0 0 0 289 247 57 9 34
2 0 0 2768 66564 176860 436628 0 0 1 229 405 336 55 23 22
3 0 0 2768 66792 176860 436628 0 0 0 0 242 145 14 59 27
2 0 0 2768 66688 176860 436628 0 0 2 0 250 160 25 62 13
2 0 0 2768 66096 176860 436636 0 0 0 0 336 226 29 56 16
2 0 0 2768 66248 176860 436640 0 0 0 0 315 222 30 54 16
2 0 0 2768 66920 176860 436640 0 0 0 0 214 104 14 56 30
1 0 0 2768 67160 176860 436624 0 0 0 0 228 138 17 52 30
3 0 0 2768 373596 176860 130336 0 0 3 0 257 127 8 58 33
3 0 0 2768 396532 176860 106860 0 0 753 0 510 321 30 59 11
6 1 0 2768 390452 176860 112028 0 0 1252 97 536 369 35 60 5
5 0 0 2768 387568 176860 113264 0 0 56 0 383 278 55 45 0
5 0 0 2768 391424 176860 110360 0 0 46 0 396 272 14 86 0
3 0 0 2768 392592 176860 110720 0 0 95 0 376 240 8 92 0
3 0 0 2768 392248 176860 110732 0 0 13 0 309 214 10 90 0
5 0 0 2768 391992 176860 110764 0 0 10 0 289 172 29 71 0
5 0 0 2768 392040 176860 110776 0 0 2 0 318 204 29 71 0
3 0 0 2768 392744 176860 110884 0 0 32 0 322 177 10 88 3
6 0 0 2768 392704 176860 110884 0 0 0 0 203 94 8 79 13
...Note the huge sudden drop in cache memory and free memory going way
up. This happened just after I started the "du". I thought maybe
something allocated and freed a whole bunch of memory really quickly,
which could explain this sudden drop, but I was curious so I tried the
"du -s *" again.
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
2 0 0 2768 213888 182608 282696 0 0 96 0 472 384 25 64 12
1 0 0 2768 214440 182608 282688 0 0 2 0 486 394 8 63 29
2 0 0 2768 214960 182608 282680 0 0 0 0 355 273 11 59 30
4 0 0 2768 348468 182608 149572 0 0 0 0 317 235 31 60 10
5 0 0 2768 346732 182608 150408 0 0 210 0 336 243 24 59 17
3 0 0 2768 418240 182608 80288 0 0 98 59 385 274 22 60 18
4 0 0 2768 417276 182608 80288 0 0 0 0 295 227 33 51 16
3 1 0 2768 414736 182608 82500 0 0 1061 0 724 357 28 60 12
5 1 0 2768 407900 182608 86996 0 0 1123 0 613 335 30 67 3
5 0 0 2768 403820 182608 90804 0 0 946 0 704 319 22 78 0
Ouch! Is the wrong stuff being freed or something, maybe?
I tried it five times now, and each time I run the "du", it throws out a
whole bunch of cache for no apparent reason. "ls -R" seems to make it do
the exact same thing.
Hmm... Looking at /proc/sys/fs/dentry-state, I see it going up and down
very quickly (going up and then freeing a whole bunch at once), and as
this happens the cache is being tossed out.
I guess this only really happens when people lookup or open a whole bunch
of different files quickly, but it might be something to look at...
Simon-
[ Stormix Technologies Inc. ][ NetNation Communications Inc. ]
[ sim@stormix.com ][ sim@netnation.com ]
[ Opinions expressed are not necessarily those of my employers. ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon May 15 2000 - 21:00:23 EST