All of lore.kernel.org
 help / color / mirror / Atom feed
From: Marc MERLIN <marc@merlins.org>
To: Roman Mamedov <rm@romanrm.net>
Cc: linux-block@vger.kernel.org, linux-raid@vger.kernel.org
Subject: Re: 5.1.21 Dell 2950 terrible swraid5 I/O performance with swraid on top of Perc 5/i raid0/jbod
Date: Mon, 19 Aug 2019 12:16:25 -0700	[thread overview]
Message-ID: <20190819191625.GH5431@merlins.org> (raw)
In-Reply-To: <20190819233709.76900bbf@natsu>

On Mon, Aug 19, 2019 at 11:37:09PM +0500, Roman Mamedov wrote:
> > > Default Cache Policy: WriteBack, ReadAheadNone, Direct, No Write Cache if Bad BBU
> > > Current Cache Policy: WriteBack, ReadAheadNone, Direct, No Write Cache if Bad BBU
> > > Default Access Policy: Read/Write
> > > Current Access Policy: Read/Write
> > > Disk Cache Policy   : Disabled
> 
> So does it have a BBU? (Battery backup unit)
 
Yes, it does and it's working.
But because I found that with write caching enabled, it seemd to take all
the writes from the raid rebuild in a big queue, and starving I/O for others
requests that I wanted to happen "right now" (like /bin/ls actually being
loaded and running), and reading how the perc 5/i is a crap card, I turned
off its IO caching, leaving the work to linux' block buffer and the 32GB of
RAM in the server that are mostly allocated to disk IO caching.

> > I tried to disable the card's write cache to let linux and its 32GB of
> > RAM, do it better, but I didn't see a real improvement:
> 
> I'd expect that on the contrary, you should look for ways to enable it, and
> force-enable even without that BBU (in case of lack of one), because it feels
> like what you did is disable disks' own write buffering, and not (only) the
> card's!

Yes, you may be correct on that. I can re-enable it, but it was terrible
with it on, too.

> What you are observing seems to me like what "dd" does with "oflag=dsync" (and
> comparable performance that it gets). Definitely feels like it's in some
> "extra safe mode" and, say, every 4KB piece of data leads to full flush to
> disk before accepting to write the next 4KB.

That sounds plausible indeed.

> More things to try, check if it's possible to set up disks not as 1-member
> RAID0, but 1-member "linear" ("JBOD"), or even 1-member RAID1, who knows maybe
> some of this would work better.

Assuming I can do this without losing the entire filesystem, I can try, but
if it can't do single drive raid0, I doubt changing this to single drive
raid1 would make things much better.
Then again, once you're hitting things that aren't working as they should...

I have an H700 in the mail that should arrive tonight, I'll try swapping
that first and see what happens.

Thanks for the answer,
Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/  

      reply	other threads:[~2019-08-19 19:16 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-19  7:08 5.1.21 Dell 2950 terrible swraid5 I/O performance with swraid on top of Perc 5/i raid0/jbod Marc MERLIN
2019-08-19  9:18 ` Paolo Valente
2019-08-19 12:02   ` Paolo Valente
2019-08-19 16:40   ` Marc MERLIN
2019-08-19 17:05     ` Paolo Valente
2019-08-19 17:26       ` Marc MERLIN
2019-08-19 11:42 ` o1bigtenor
2019-08-19 16:24   ` Marc MERLIN
2019-08-20  5:49   ` Marc MERLIN
2019-08-19 18:37 ` Roman Mamedov
2019-08-19 19:16   ` Marc MERLIN [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190819191625.GH5431@merlins.org \
    --to=marc@merlins.org \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-raid@vger.kernel.org \
    --cc=rm@romanrm.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.