Date: Wed, 1 Oct 2008 13:02:11 +0200 From: =?ISO-8859-1?Q?Johan_Str=F6m?= <johan@stromnet.se> To: Danny Braniss <danny@cs.huji.ac.il> Cc: freebsd-scsi@freebsd.org Subject: Re: state of iscsi_initiator? Message-ID: <A7673452-0862-421F-B07B-CFDEA2C7009A@stromnet.se> In-Reply-To: <E1KkuNM-0005Kr-PX@cs1.cs.huji.ac.il> References: <35FAB782-F576-4FB7-8633-67CFB63D9389@stromnet.se> <48E300CB.8070808@modulus.org> <E1KkuNM-0005Kr-PX@cs1.cs.huji.ac.il>
next in thread | previous in thread | raw e-mail | index | archive | help
On Oct 1, 2008, at 7:35 AM, Danny Braniss wrote: >> >> I'm just playing with iSCSI now, on a fresh 7.1-STABLE machine >> connected >> to another FreeBSD box running ports/net/iscsi-target. >> >> I am using the latest version 2.1 Initiator from: >> ftp://ftp.cs.huji.ac.il/users/danny/freebsd/ >> >> It seems alot better now than when I last tried it. The speed is >> constant at around 50MB/s for a single sequential read or write. It >> also reconnected very quickly when I restarted the target. I was >> even >> able to reboot the target box and I/O resumed perfectly once it >> finished >> booting! >> >> >> The speed increased to 80mb/s when I ran two threads, which is >> close to >> the max performance for the network cards in this machine. Nice, I guess. Wonder how much one can get up to with 1GBit.. Altough then I guess thats when the disks will be the limiting factor. As for recovery, that have been working pretty nice for me to, I've been running 2 VMWare machines here (on the same physcal box), one with OpenFiler and one with 7.0-STABLE (and 2.0.99 iscsi). I've been running iozone for a couple of hours, without any crashes or so at least. I also tried rebooting the openfiler in the middle of the iozone process with the following results (where the numbers are iozone output): 2048 128 2111 3961 2547261 2190626 2048 256 4405 4363 2443638 2233342 2048 512iscontrol[3687]: trapped signal 30 trapped signal 30 iscontrol: supervise going down iscontrol[3687]: sess flags=2000040d iscontrol[3687]: Reconnect errno=61 connect: Connection refused iscontrol[3687]: Reconnect iscontrol: supervise starting main loop 21 4271 1391257 2098062 2048 1024 4053 4151 2389263 4113514 2048 2048 3634 5330 2233342 3475985 4096 4 2860 3947 1648862 1836000 Yes iscontrol was started from the same console earlier. Regarding that, I noticed that the 2.1 version had a daemonized version. is there any changelog for 2.1? >> >> >> >> A couple of minor issues: >> >> 1. Trouble shutting down iscontrol - after I kill -9 it stays >> around >> for a minute or more until it finally goes away. >> > > you should kill -1, this does an orderly shutdown of the connection. > I guess an update to the manual is needed. > >> 2. Under very heavy multi-threaded load, the client machine gets >> sluggish to respond, even to keystrokes on the console. And if I >> shutdown the target, even local disk accesses on the client computer >> stop working after a while. If you take the target back up, does it wake up again? >> >> >> (I don't think this is a problem with iSCSI initiator. I suspect >> there's something in FreeBSD which pauses all disk I/O if the >> length of >> outstanding requests is too high, can anyone shed some light on >> this?) >> > I've noticed slugishness when a heavy fsck is running in the > background - not > iscsi related, so it seems that heavy disk IO is using up resources, > but > then again, the CPU is working :-) Thanks for your input :)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?A7673452-0862-421F-B07B-CFDEA2C7009A>