Re: [Nbd] Root-on-NBD
- To: Wouter Verhelst <w@...112...>
- Cc: firstname.lastname@example.org
- Subject: Re: [Nbd] Root-on-NBD
- From: dsuchod <dsuchod@...65...>
- Date: Fri, 17 Jul 2009 23:42:10 +0200 (CEST)
- Message-id: <alpine.LNX.2.00.0907172338100.12707@...66...>
- In-reply-to: <20090717153305.GB4033@...510...>
- References: <20090717133935.GA32065@...510...> <4A609801.8050303@...124...> <20090717153305.GB4033@...510...>
Sounds reasonable to me. I'm not sure I understand the need to
disconnect though. What is the issue that occurs if you don't
Just that the server doesn't notice the client is gone until the TCP
keepalive probes time the connection out. It shouldn't be a major issue,
but on resource-strapped servers that serve hundreds of clients, it
might be better to avoid it if possible. Also, it's just cleaner to
properly disconnect if we know we're going down anyway, and I don't see
any downside to doing this (but then, I could be missing something).
Sounds good! We are using NBD in combination with SQFS on diskless
workstations. Often the list of pending server threads does not give
the full picture of connected clients because they are dead already.
So it would be nice to have it!
A bit related question: Would it be possible to survive a suspend-to-
ram with NBD? Problem would be the dead TCP session after a while ...