this suggests that turning off IPv6 might increase performance.
So I tried it. And in a gigE configuration I could not find a difference.
with IPv6:
PowerBook G4 A15 SGI Tezro
[ 3] 0.0-10.0 sec 863 MBytes 724 Mbits/sec
[ 3] 0.0-10.0 sec 760 MBytes 638 Mbits/sec
[ 3] 0.0-10.0 sec 708 MBytes 594 Mbits/sec
PowerBook G4 A15 Linux x86
[ 3] 0.0-10.2 sec 484 MBytes 397 Mbits/sec
[ 3] 0.0-10.0 sec 584 MBytes 490 Mbits/sec
[ 3] 0.0-10.2 sec 568 MBytes 467 Mbits/sec
without IPv6:
PowerBook G4 A15 SGI Tezro
[ 3] 0.0-10.0 sec 788 MBytes 661 Mbits/sec
[ 3] 0.0-10.0 sec 744 MBytes 624 Mbits/sec
[ 3] 0.0-10 0 sec 872 MBytes 731 Mbits/sec
PowerBook G4 A15 Linux x86
[ 3] 0.0-10.0 sec 581 MBytes 487 Mbits/sec
[ 3] 0.0-10.2 sec 608 MBytes 500 Mbits/sec
[ 3] 0.0-10.0 sec 598 MBytes 502 Mbits/sec
Your Mileage May Vary
An easy way to run things on remote machines in unix has been the ‘rsh’ command.
For instance:
rsh -l guest remote-machine command-to-run-on-that-machine
would execute the command on the other machine. Of course things need to be set up for that. It used to be simple, but the default installs
for redhat redhat (and possible other installs) need a bit of tweaking. Here a quick and dirty recipe. This openes the machine to the public,
so don’t do it on any computer exposed to the internet or a bunch of untrustworthy users.
useradd guess
(remove !! from /etc/shadow for user gues)
A user guest with no password exists. Cool and scary. Up to you.
rpm -i rsh-server-0.17-21.i386.rpm
Versions might be different, check if it does already exist
(set disabled="off" to disabled="on" in /etc/xinetd.d/rsh)
killall -HUP xinetd
Tell xinetd that it is allowed to start rshd if an incoming connection is being made.
cp /etc/pam.d/login /etc/pam.d/rsh
This is a true hack for people that don't know what they are doing. I have no idea what
the side effects are of this. It worked for what I had to do, but it might be completely
inapproiate.
From now on you can do things on that machine as the user 'guest'.
Citigroup looses tapes with the records
of 3.7 million clients. What _would_ have be cool would be if they would have not have copies.
That would teach them.
Right now all those 3.7 million people can hope for is that UPS lost the tapes real good. They can loose things so good that nobody can find them.
Really, only fedex and UPS can do that.
Finally the stack overflow exploits handcrafted for intel CPU might start work on Macs!
malware alliance
What would happen if the masses of recruited Windows PCs are able to impact bigger part
of the internets, so that outages will be noticebale for more people?
Think “Dr. Evil”. :
You want the internet back? That would be “one million dollars”.
Thanks Microsoft. I hope Bill is paying the ransom he and his OS have caused.
It’s not the internet that is vunerable, it is not the computers. It is the operating system called Windows made by Microsoft.
Technically all systems can have viruses. In reality only Microsoft Windows systems are part of these malware empires.
Since people tend to say different: This has nothing to do with market share. 10% Apple Systems is by far enough to be
attractive. In the webserver market Mirocroft products are the minority but still manage to host all the interesting exploits.
It’s a design problem, and a historical one. For Windows security the geenie is out of the bottle. Apple can afford to fix every problem that becomes known: Their virus count is zero. It is so much easier to go back to zero from one than from multiple thousand.
a brief history of the internet
Linked from there: Bill Gates on “Hobby Software” 1976.
Take a box, put a harddrive in it, connect it to the internet to get video content, connect it to the TV, sell it for 100 US$.
Sounds like a good plan these days. It’s called “Akimbo”, and the New York Times does not like it at all. And I can understand why. They have two thousand titles. Which really is nothing. It reminds me of those days when people scraped ‘the best of the internet’ on a server and put it on a plane.
Looks like the AP did a Akimbo ‘review’ a couple of days ago.
Same results.
Just finishing listening to On A Note of Triumph on NPR. I am not able to recommend it vividly enough.
Sometimes I wonder if I would still be in LA if there would be no KCRW.
It all went to well.
suddenly the server was not reacting. Everything worked till the shell needed to do /anything/ with the disk.
Had to reset it. Sigh.
The syslog said:
May 29 19:14:25 andreaswacker kernel: kswapd0: page allocation failure. order:5, mode:0x50
May 29 19:14:25 andreaswacker kernel: [<0214b29a>] __alloc_pages+0x28b/0x298
May 29 19:14:25 andreaswacker kernel: [<0214b2bf>] __get_free_pages+0x18/0x24
May 29 19:14:25 andreaswacker kernel: [<0214e9c1>] kmem_getpages+0x15/0x94
May 29 19:14:25 andreaswacker kernel: [<0214f74c>] cache_grow+0x155/0x29a
May 29 19:14:25 andreaswacker kernel: [<0214fa9e>] cache_alloc_refill+0x20d/0x23d
May 29 19:14:25 andreaswacker kernel: [<0215004f>] __kmalloc+0x6b/0x7d
May 29 19:14:25 andreaswacker kernel: [<82964f54>] kmem_alloc+0x50/0x96 [xfs]
May 29 19:14:25 andreaswacker kernel: [<829479a1>] xfs_inode_item_format+0xe0/0x239 [xfs]
May 29 19:14:25 andreaswacker kernel: [<8295aaf3>] xfs_trans_fill_vecs+0x3a/0x86 [xfs]
May 29 19:14:25 andreaswacker kernel: [<8295a8a4>] xfs_trans_commit+0x18d/0x300 [xfs]
May 29 19:14:25 andreaswacker kernel: [<829494de>] xfs_iomap_write_allocate+0x248/0x436 [xfs]
May 29 19:14:25 andreaswacker kernel: [<82949520>] xfs_iomap_write_allocate+0x28a/0x436 [xfs]
May 29 19:14:25 andreaswacker kernel: [<0224e474>] generic_make_request+0x190/0x1a0
May 29 19:14:25 andreaswacker kernel: [<829485a3>] xfs_iomap+0x23b/0x3ed [xfs]
May 29 19:14:25 andreaswacker kernel: [<829486ba>] xfs_iomap+0x352/0x3ed [xfs]
May 29 19:14:25 andreaswacker kernel: [<8296c99f>] xfs_bmap+0x1a/0x1e [xfs]
May 29 19:14:25 andreaswacker kernel: [<82965219>] xfs_map_blocks+0x29/0x11e [xfs]
May 29 19:14:25 andreaswacker kernel: [<82965dd9>] xfs_page_state_convert+0x273/0x4e8 [xfs]
May 29 19:14:25 andreaswacker kernel: [<829664cb>] linvfs_writepage+0x91/0xc6 [xfs]
May 29 19:14:25 andreaswacker kernel: [<021522db>] pageout+0x83/0xc0
May 29 19:14:25 andreaswacker kernel: [<02152522>] shrink_list+0x20a/0x547
May 29 19:14:25 andreaswacker kernel: [<02151540>] __pagevec_release+0x15/0x1d
May 29 19:14:25 andreaswacker kernel: [<02152a92>] shrink_cache+0x233/0x4d5
May 29 19:14:25 andreaswacker kernel: [<0215357f>] shrink_zone+0x8f/0x9a
May 29 19:14:25 andreaswacker kernel: [<021538a5>] balance_pgdat+0x176/0x249
May 29 19:14:25 andreaswacker kernel: [<02153a3e>] kswapd+0xc6/0xc8
May 29 19:14:25 andreaswacker kernel: [<02120fcf>] autoremove_wake_function+0x0/0x2d
May 29 19:14:25 andreaswacker kernel: [<02120fcf>] autoremove_wake_function+0x0/0x2d
May 29 19:14:25 andreaswacker kernel: [<02153978>] kswapd+0x0/0xc8
May 29 19:14:25 andreaswacker kernel: [<021041d9>] kernel_thread_helper+0x5/0xb
May 29 19:14:25 andreaswacker kernel: deadlock in kmem_alloc (mode:0x50)
May 29 19:14:25 andreaswacker kernel: possible deadlock in kmem_alloc (mode:0x50)
May 29 19:14:25 andreaswacker last message repeated 85 times
Don’t hope that this is common for Fedora Core3 on a AMD machine with a big array. I start to load the machine with tasks now.
Let’s see if it happens again.
Sure enough Mysql was not happy since I run it with delay-key-write.
WordPress said:
WordPress database error: [Incorrect key file for table 'wp_comments'; try to repair it]
So I did a
mysqlcheck -pXXXXXX --auto-repair wordpress
which seems to have done the trick.