xfs
[Top] [All Lists]

Re: xfs logbufs default, (also noatime, lazy-count(of partition formatte

To: "Linda A. Walsh" <xfs@xxxxxxxxx>
Subject: Re: xfs logbufs default, (also noatime, lazy-count(of partition formatter)...32767b v. 32768b...
From: Eric Sandeen <sandeen@xxxxxxxxxxx>
Date: Tue, 25 Aug 2009 08:52:36 -0700
Cc: xfs-oss <xfs@xxxxxxxxxxx>
In-reply-to: <4A92E922.2030404@xxxxxxxxx>
References: <4A92E922.2030404@xxxxxxxxx>
User-agent: Thunderbird 2.0.0.23 (Macintosh/20090812)
Linda A. Walsh wrote:
Is there a reason why logbufs=8 should't become the default?
Isn't it the best choice for most users given today's memory
constraints (or lack thereof?)

It is the default:

#define XLOG_MAX_ICLOGS         8

        if (mp->m_logbufs <= 0)
                log->l_iclog_bufs = XLOG_MAX_ICLOGS;


I saw a message about xfs parameter tuning and still see the mention to:
'logbufs=8,noatime' having to be specified on every mount.

you can't believe everything you read on the internets ;)

If the desire is to provide default optimal tuning, maybe a default of logbufs to 8 would be a good idea, and maybe a
mkfs flag to default to 'noatime' and a 64MB internal log
size would be good ideas for *most* users?

noatime is something a user should choose based on their needs, not a default. Some things depend on atime, and noatime isn't really posix-compliant...

Log size depends on the fs size...

How much of problem is defaulting to a 64MB internal log
size when file partitions are measured in the 100GB range?

                /*
                 * With a 2GB max log size, default to maximum size
                 * at 4TB. This keeps the same ratio from the older
                 * max log size of 128M at 256GB fs size. IOWs,
                 * the ratio of fs size to log size is 2048:1.
                 */

100*1024/2048
50

so it currently scales to 50M for 100G.

I'm not sure how much use it is to even remember typing lazy-count=1
on the size partition each time.

lazy-count will/should be the default soon; someone (me) just needs to make xfstests cope with the change first.

One more 'nit', while I think of it...for a 64MB log buffer size,
one should be able to specify 32768b, but last I tried, that gave
a 'too large' error...isn't 32768*4=64MB, not 32767b?

seems fine?

$ /sbin/mkfs.xfs -dfile,name=fsfile,size=100g -lsize=32768b
meta-data=fsfile                isize=256    agcount=4, agsize=6553600 blks
         =                      sectsz=512   attr=2
data     =                      bsize=4096   blocks=26214400, imaxpct=25
         =                      sunit=0      swidth=0 blks
naming   =version 2             bsize=4096   ascii-ci=0
log      =internal log          bsize=4096   blocks=32768, version=2
         =                      sectsz=512   sunit=0 blks, lazy-count=0
realtime =none                  extsz=4096   blocks=0, rtextents=0

-Eric

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