> spinlocks
I first heard the term SpinLock when ADP Dealer Services had migrated from
their pre-UNIX OS to UNIX. It took them years to get that resolved. Once
they cut to Red Hat Linux the problem had by then been rpetty much brought
down to a fairly benign level, and finally they pretty much nailed it. I
have yet to hear how they did eventually resolve the problem, but I suspect
it had a lot to do with record locking/unlocking with their PICK database
(non)solution from 30+ years ago...
Gil
> -----Original Message-----
> From: prolinux-bounces AT leafe D.OT com [mailto:prolinux-bounces@leafe.com]On
> Behalf Of Paul McNett
> Sent: Thursday, February 28, 2008 4:03 PM
> To: ProLinux Email List
> Subject: Re: [ProLinux] Samba, oplocks, spinlocks and lock spin time?
>
>
> By the way, I've never used or even heard of 'spinlock'. I only ever set
> oplocks, level 2 oplocks, and fake oplocks to 'no'.
>
> Also, I never have to reindex the tables for this app. I do it anyway
> every few months, but never because it becomes necessary. There are
> about 70 tables, more views than that, and 20+ users. I open tables as
> needed, but once opened I never close them. I actually only deal with
> views in my code.
>
> I mention this only as a case of samba working really well, so you know
> it is possible! I switched to Linux/Samba because I was tired of
> reindexing all the time, and having to reboot the server for whatever
> stupid reason[1]. It was the best decision I ever made for this client:
> I was getting really worried that my app wasn't scaling up as they kept
> adding users, that I should have gone client/server to begin with. I was
> glad to find it wasn't my app but the server platform in use.
>
> Now, we are working on making it client/server (and rewriting in Dabo),
> but I didn't have to push them to pay for the transition 5 years ago,
> which they weren't in a position to do then.
>
> Paul
>
> [1] uptime for this server executed just now:
> 12:59:34 up 215 days, 22:14, 2 users, load average 0.00, 0.00, 0.00
>
> 215 days ago, the electric company needed to replace a transformer.
> Before that, I believe we were seeing uptimes of over 100 days, with
> reboots for similar reasons not having to do with the server software.
>
> I used to restart samba in the middle of the night, but then found that
> it wasn't necessary and it had the bad effect of killing peoples mapped
> drives if they happened to forget to log off at the end of the day. It
> also would mess things up horribly if they happened to have my vfp app
> open at the point samba was restarted. Duh.
>
>
[excessive quoting removed by server]
©2008 Gil Hale |
