[prev in list] [next in list] [prev in thread] [next in thread] 

List:       opensolaris-storage-discuss
Subject:    Re: [storage-discuss] Offlining a LU timing
From:       "Steffen Plotner" <swplotner () amherst ! edu>
Date:       2009-12-18 20:57:18
Message-ID: 0DA55A84B13D704AAC5C23A94F3769EC81E470 () mail8 ! amherst ! edu
[Download RAW message or body]

> -----Original Message-----
> From: Peter.Cudhea@Sun.COM [mailto:Peter.Cudhea@Sun.COM] 
> Sent: Wednesday, December 16, 2009 6:05 PM
> To: Sumit Gupta
> Cc: Steffen Plotner; storage-discuss@opensolaris.org
> Subject: Re: [storage-discuss] Offlining a LU timing
> 
> At a first glance, it seems like a reasonable symptom given that the 
> network interfaces have been removed.   The threads that are 
> waiting in 
> iscsit_abort appear to have commenced sending a SCSI response PDU and 
> have not yet received confirmation that the PDU has been sent out the 
> door.  I am not sure what causes it to clear after 20 
> seconds.   Perhaps 
> the underlying TCP connection goes down at that time.  Do you by any 
> chance have a threadlist from after the LU has finished being brought 
> offline?   Do the LUs go offline faster if you skip the step 
> of bringing 
> the network interfaces down first?

Yes, offlining the LUs while the network interfaces are speeds up the
offlining process. I was watching the output of netstat to see the
connections drop much faster while the network interfaces were up.

> 
> The following thread stack trace shows that TCP traffic is blocked 
> getting out the network interface, i.e. that the TCP send 
> queue is full. 
> That is the case in which so_snd_wait_qnotfull_locked will 


Thank you for your response.

Steffen
_______________________________________________
storage-discuss mailing list
storage-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
[prev in list] [next in list] [prev in thread] [next in thread]