Re: kernel work week of 3-Feb-2010 HEADS UP
--00032555af929bd084047ed81b62
Content-Type: text/plain; charset=ISO-8859-1
2010/2/5 Matthew Dillon <dillon@apollo.backplane.com>
> :Btw, that should be already possible using multiple HAMMER volumes.
> :The SSD partition should be made the root volume. It can be very
> :small, so that only the UNDO log fits on it (maybe a GB?), the second
> volume
> :would then be the regular hard disk. Maybe we'd need to give newfs_hammer
> :a specific option so that it treats all space of the first volume as UNDO
> :and
> :uses the second volume for storage.
> :
> :I think I can implement that. Matt, do you think the option to
> newfs_hammer
> :is
> :a good idea?
> :
> :Regards,
> :
> : Michael
>
> Not for newfs_hammer. It would be a good fit for the hammer
> volume-add directive, though.
>
Okay, I was just thinking about the lowest hanging fruit, and changing
a few lines in newfs_hammer (basically just the space calculations)
seemed to be those fruits :)
>
> You could add a feature where it adds the volume and moves the
> UNDO FIFO into it. volume-del might also have to be adjusted to
> move the UNDO FIFO back out (I don't recall offhand if it moves
> UNDO blocks).
>
It doesn't move UNDO yet. Right now, the UNDO log is always on
the root volume, but when we add another volume, we allocate the
same space on the new volume for the UNDO log. So at least we
are well prepared for moving the UNDO.
Your idea sounds good, and would lead us to one step closer to
root volume removal.
Could I implement moving the UNDO log by flushing everything to disk, so
that the UNDO log gets empty and then simply switching the log to the
new one? IIRC, writing an UNDO entry required the root volume to be
updated in HAMMER versions <= 3, which is no longer true as of V4.
Regards,
Michael
--00032555af929bd084047ed81b62
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
<br><br><div class=3D"gmail_quote">2010/2/5 Matthew Dillon <span dir=3D"ltr=
"><<a href=3D"mailto:dillon@apollo.backplane.com">dillon@apollo.backplan=
e.com</a>></span><br><blockquote class=3D"gmail_quote" style=3D"margin:0=
0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
:Btw, that should be already possible using multiple HAMMER volumes.<br>
<div class=3D"im">:The SSD partition should be made the root volume. It can=
be very<br>
:small, so that only the UNDO log fits on it (maybe a GB?), the second volu=
me<br>
:would then be the regular hard disk. Maybe we'd need to give newfs_ham=
mer<br>
:a specific option so that it treats all space of the first volume as UNDO<=
br>
:and<br>
:uses the second volume for storage.<br>
:<br>
:I think I can implement that. Matt, do you think the option to newfs_hamme=
r<br>
:is<br>
:a good idea?<br>
:<br>
:Regards,<br>
:<br>
: =A0Michael<br>
<br>
</div> =A0 =A0Not for newfs_hammer. =A0It would be a good fit for the hamme=
r<br>
=A0 =A0volume-add directive, though.<br></blockquote><div><br></div><div>O=
kay, I was just thinking about the lowest hanging fruit, and changing</div>=
<div>a few lines in newfs_hammer (basically just the space calculations)=A0=
</div>
<div>seemed to be those fruits :)</div><div><br></div><div>=A0</div><blockq=
uote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc =
solid;padding-left:1ex;">
<br>=A0=A0 You could add a feature where it adds the volume and moves the<b=
r>
=A0 =A0UNDO FIFO into it. =A0volume-del might also have to be adjusted to<=
br>
=A0 =A0move the UNDO FIFO back out (I don't recall offhand if it moves=
<br>
=A0 =A0UNDO blocks).<br></blockquote><div><br></div><div>It doesn't mo=
ve UNDO yet. Right now, the UNDO log is always on</div><div>the root volume=
, but when we add another volume, we allocate the</div><div>same space on t=
he new volume for the UNDO log. So at least we=A0</div>
<div>are well prepared for moving the UNDO.</div><div><br></div><div>Your i=
dea sounds good, and would lead us to one step closer to</div><div>root vol=
ume removal.</div><div><br></div><div>Could I implement moving the UNDO log=
by flushing everything to disk, so</div>
<div>that the UNDO log gets empty and then simply switching the log to the<=
/div><div>new one? IIRC, writing an UNDO entry required the root volume to =
be</div><div>updated in HAMMER versions <=3D 3, which is no longer true =
as of V4.</div>
<div><br></div><div>Regards,</div><div><br></div><div>=A0=A0Michael</div><d=
iv><br></div></div>
--00032555af929bd084047ed81b62--
討論串 (同標題文章)
完整討論串 (本文為第 21 之 30 篇)：