xfs
[Top] [All Lists]

Re: [PATCH 2/2] xfs: mark the xfs-alloc workqueue as high priority

To: Eric Sandeen <sandeen@xxxxxxxxxxx>
Subject: Re: [PATCH 2/2] xfs: mark the xfs-alloc workqueue as high priority
From: Tejun Heo <tj@xxxxxxxxxx>
Date: Tue, 13 Jan 2015 15:46:33 -0500
Cc: Eric Sandeen <sandeen@xxxxxxxxxx>, xfs-oss <xfs@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=sender:date:from:to:cc:subject:message-id:references:mime-version :content-type:content-disposition:in-reply-to:user-agent; bh=0kNxCKm95qGpsczlNZDvdDZiM6uY8J7WS9wz4HYBmuc=; b=Pq7znYvutOka52/2cIhLbgfwYmx2HuiDnwE7kIVB306gcSklf8Xgv+l9W67QbJmMio AibKD+oO2s8Z4MrPOy7tWWE6Jdf+A1OzIfanPKmd0F44dG4MlQ2hdmkvnoB+jJjc1Fn5 imlDmj5hV2iTUMLy1I/s2x09Sz7bHqJLto+wI3nVZK5v6EH5Nb4qw8tIvqwG7/G91DDF fDCHB/3xWhYtnUghJPjx1OG4NFAIQXwry8qlHZSHIsDhDL+4NpGhmlSwhJQry9zCg0WT IHWahE36pXJtB5kkmp8af1KbeZxcAdOZ0nM7NTRP93NSvp3NgdiBBgFC+Q0rZqHr7pln hUKQ==
In-reply-to: <54B58041.9070502@xxxxxxxxxxx>
References: <20150109182310.GA2785@xxxxxxxxxxxxxx> <54B03BCC.7040207@xxxxxxxxxxx> <20150110192852.GD25319@xxxxxxxxxxxxxx> <54B429EB.9050807@xxxxxxxxxxx> <20150112225314.GC22156@xxxxxxxxxxxxxx> <54B454E2.70707@xxxxxxxxxxx> <20150112233755.GD22156@xxxxxxxxxxxxxx> <54B56D2B.6090401@xxxxxxxxxxx> <20150113201900.GA9489@xxxxxxxxxxxxxx> <54B58041.9070502@xxxxxxxxxxx>
Sender: Tejun Heo <htejun@xxxxxxxxx>
User-agent: Mutt/1.5.23 (2014-03-12)
Hello, Eric.

On Tue, Jan 13, 2015 at 02:29:53PM -0600, Eric Sandeen wrote:
> > Can you please also report the value of nr_running?  That's what
> > regulates the kick off of new workers and rescuers.
> 
> sorry about that, swapped nr_workers w/ nr_running in my brain:
> 
>   nr_running = {
>     counter = 0
>   }, 

So,

        nr_workers == 15,
        nr_idle == 0,
        nr_running == 0,

That means one worker must be playing the role of manager by executing
manage_workers() whic his also responsible for kicking off the
rescuers if it fails to create new workers in a short period of time.
The manager is identifier as the holder of pool->manager_arb and while
a manager is trying to creat a worker, pool->mayday_timer must be
armed continuously firing off every MAYDAY_INTERVAL summoning rescuers
to the pool, which should be visible through the pool_pwq->mayday_node
corresponding to the stalled pool being queued on wq->maydays.

Can you post the full dump of the pool, wq and all kworkers?

Thanks.

-- 
tejun

<Prev in Thread] Current Thread [Next in Thread>