1 Reply Latest reply on Dec 13, 2012 1:55 PM by 854078

    Solaris 11.1 AI client: solaris.zlib download times out

      Hi all,

      I am trying to install a client off the current AI image. I have not modified any manifests and just run the default config after "installadm create-service" which should give me an interactive text install mode. The client has the Intel I350-T2 adapter ( [http://ark.intel.com/products/59062/Intel-Ethernet-Server-Adapter-I350-T2] ) installed. The PXE part finishes quite fine, the client receives the IP from the DHCP server, the boot loader and the subsequent boot_archive download quite fine over TFTP.

      But just after the Solaris installer loads ("Probing for device nodes ..."), the connectivity is lost (ICMP echo requests are not answered anymore) and subsequently, the HTTP download of the solaris.zlib network image fails ("Connecting to: <IP address>:5555... failed: Connection timed out.") . I would suspect that the NIC driver's probing simply is unable to recognize the i350 network adapter, but when I can drop to the shell (after 10 failed connection retries), I can run "ifconfig -a" and see the "igb0" interface plumbed and configured with DHCP (at least for a while until the lease expires) - so something appears to be detected by the driver and configured according to the lease granted to the host in the PXE phase. /system/volatile/install_log does not exist at that point, /var/log/syslog is empty and the command set is very limited (no "dmesg"), so I am a bit stuck without any usable log files available.

      Am I guessing correctly that I would need a specific NIC driver set upon install for the i350? If so, how would I go about integrating an additional NIC driver into the AI install procedure?

      Thanks, kind regards,


      Edited by: user13673072 on 11.12.2012 06:18
        • 1. Re: Solaris 11.1 AI client: solaris.zlib download times out
          After having performed a manual installation from media, I have noticed problems with the Ethernet interfaces on this system - packets were timing out, connections seemed dropped. A look into "fmadm faulty" presented me with a couple of events for pci@0,0/pci8086,3c08@3/pci15d9,1521. According to "prtconf -v" it is a "pciclass 0200" (Network Controller) and is carrying a "compatible" alias of pci8086,1521 (Intel i350 Powerville NIC). I double-checked with other OSes and the NICs seem running quite fine on the same machine, so the problem is apparently with the hardware's driver.

          At least I am sufficiently positive now that it is not an AI issue, thanks for reading.