Hi Mike and List, Thanks for the reply.
On Wed, May 21, 2008 at 12:28 AM, Mike Christie <[EMAIL PROTECTED]> wrote: > > Shrey wrote: >> Hi Mike, >> >> Thanks for you reply. Please find my reply inline. >> >> On May 19, 11:04 pm, Mike Christie <[EMAIL PROTECTED]> wrote: >>> Shrey wrote: >>>> Dear All, >>>> I started using/studying open-iscsi recently. I was wondering if there >>>> is any hard-coded limit to number of targets devices that can >>>> discovered per iSCSI target? Is there any hard-coded limit on the >>>> number of LUNs supported per target device (assuming that somehow HBA >>>> on iscsi target doesn't present any limit) ? >>> As others pointed out there is a module param for the max lun for >>> iscsi_tcp, but that should actually be limited by a libiscsi/scsi-ml >>> limit. libiscsi uses a function named int_to_scsilun in >>> drivers/scsi/scsi_scan.c and as you can see that only implments 2 >>> levels. libiscsi/iscsi_tcp also does not do the device scanning (the >>> iscsi layer only finds the targets) so we are limited by scsi_scan.c. To >>> get the limits for that do "modinfo scsi_mod". >> >> Ok, that gives me something to thing/experiment about. Thanks >> >>> There is also a limit on the number of targets. Because we allocate a >>> scsi_host per session, and the scsi layer uses a unsigned short for the >>> host number the number of targets is 2^16. >> >> That would mean 65536 devices. But I get stuck far below that - just >> about 90 targets. > > If you are using 869.* then the problem is most likely on the IET side. > Like I said before I think you need to upgrade. I had the same problem > when I was testing my fixes for lots of targets support. > > Oh yeah, I said this in some other mail, but will say it here so people > can search for it. There is another target limit and that is the number > of files a process can have open. We open a tcp socket for each session > and we do a session to each target portal. We also will open other files > so the exact number is not known. I think it is probably around 4000. I > do not have a setup that supports that many targets yet to know exactly > though. Let us know what you find out. Hmm, that sounds quite a large value as against what I am experiencing. I will try and find out what is happening in my case - probably the iscsi-target is acting like some kind of bottleneck (the max size of PDU is my initial guess) - or may be I am missing something. Nevetheless, thanks for the info - i really appreciate your help. > >> Are some parameters which I am missing which I need to configure or >> take care of? >> I tried tinkering "node.conn.iscsi.MaxRecvDataSegmentLength = >> 65536" but there was no difference. This actually is making me think >> that target is acting as a bottleneck. >> > > That is just for normal sessions. You would want to set > discovery.sendtargets.iscsi.MaxRecvDataSegmentLength Oops - I should have noticed. Thanks once again. I will try tinkering with this. > > but like I said that will not completely fix the problem. Well, this should give me something very good to start with. I will post whatever my finding are (as soon as I find something ;) ). Thanks all. Shreyansh --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "open-iscsi" group. To post to this group, send email to email@example.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/open-iscsi -~----------~----~----~----~------~----~------~--~---