All of lore.kernel.org
 help / color / mirror / Atom feed
diff for duplicates of <4FDF9EBE.2030809@shiftmail.org>

diff --git a/a/1.1.hdr b/N1/1.1.hdr
index 2900622..5015eee 100644
--- a/a/1.1.hdr
+++ b/N1/1.1.hdr
@@ -1,2 +1,2 @@
 Content-type: text/plain; charset=UTF-8; format=flowed
-Content-Transfer-Encoding: 7BIT
+Content-transfer-encoding: 7BIT
diff --git a/a/1.2.hdr b/N1/1.2.hdr
index 24ced0b..111a9a9 100644
--- a/a/1.2.hdr
+++ b/N1/1.2.hdr
@@ -1,3 +1,2 @@
 Content-type: text/html; charset=UTF-8
-Content-Transfer-Encoding: quoted-printable
-X-MIME-Autoconverted: from 8bit to quoted-printable by oss.sgi.com id q5ILXgsG243846
+Content-transfer-encoding: 8BIT
diff --git a/a/2.hdr b/N1/2.hdr
index 18d6893..4b86001 100644
--- a/a/2.hdr
+++ b/N1/2.hdr
@@ -1,4 +1,4 @@
 Content-Type: text/plain; charset="us-ascii"
 MIME-Version: 1.0
-Content-Disposition: inline
 Content-Transfer-Encoding: 7bit
+Content-Disposition: inline
diff --git a/a/content_digest b/N1/content_digest
index 3bf8c63..6c07e6f 100644
--- a/a/content_digest
+++ b/N1/content_digest
@@ -78,88 +78,88 @@
  S.
  "\02:1.2\0"
  "b\0"
- "<html>\r\n"
- "  <head>\r\n"
- "    <meta http-equiv=\"content-type\" content=\"text/html; charset=UTF-8\">\r\n"
- "  </head>\r\n"
- "  <body bgcolor=\"#FFFFFF\" text=\"#000000\">\r\n"
- "    Hello all<br>\r\n"
- "    I am doing some testing of dm-thin on kernel 3.4.2 and latest lvm\r\n"
- "    from source (the rest is Ubuntu Precise 12.04).<br>\r\n"
- "    There are a few problems with ext4 and (different ones with) xfs<br>\r\n"
- "    <br>\r\n"
- "    I am doing this:<br>\r\n"
- "    dd if=/dev/zero of=zeroes bs=1M count=1000 conv=fsync<br>\r\n"
- "    lvs<br>\r\n"
- "    rm zeroes #optional<br>\r\n"
- "    dd if=/dev/zero of=zeroes bs=1M count=1000 conv=fsync\302\240 #again<br>\r\n"
- "    lvs<br>\r\n"
- "    rm zeroes #optional<br>\r\n"
- "    ...<br>\r\n"
- "    dd if=/dev/zero of=zeroes bs=1M count=1000 conv=fsync\302\240 #again<br>\r\n"
- "    lvs<br>\r\n"
- "    rm zeroes<br>\r\n"
- "    fstrim /mnt/mountpoint<br>\r\n"
- "    lvs<br>\r\n"
- "    <br>\r\n"
- "    On ext4 the problem is that it always reallocates blocks at\r\n"
- "    different places, so you can see from lvs that space occupation in\r\n"
- "    the pool and thinlv increases at each iteration of dd, again and\r\n"
- "    again, until it has allocated the whole thin device (really 100% of\r\n"
- "    it). And this is true regardless of me doing rm or not between one\r\n"
- "    dd and the other.<br>\r\n"
- "    The other problem is that by doing this, ext4 always gets the worst\r\n"
- "    performance from thinp, about 140MB/sec on my system, because it is\r\n"
- "    constantly allocating blocks, instead of 350MB/sec which should have\r\n"
- "    been with my system if it used already allocated regions (see below\r\n"
- "    compared to xfs). I am on an MD raid-5 of 5 hdds.<br>\r\n"
- "    I could suggest to add a \"thinp mode\" mount option to ext4 affecting\r\n"
- "    the allocator, so that it tries to reallocate recently used and\r\n"
- "    freed areas and not constantly new areas. Note that mount -o discard\r\n"
- "    does work and prevents allocation bloating, but it still always gets\r\n"
- "    the worst write performances from thinp. Alternatively thinp could\r\n"
- "    be improved so that block allocation is fast :-P (*) <br>\r\n"
- "    However, good news is that fstrim works correctly on ext4, and is\r\n"
- "    able to drop all space allocated by all dd's. Also mount -o discard\r\n"
- "    works.<br>\r\n"
- "    <br>\r\n"
- "    On xfs there is a different problem.<br>\r\n"
- "    Xfs apparently correctly re-uses the same blocks so that after the\r\n"
- "    first write at 140MB/sec, subsequent overwrites of the same file are\r\n"
- "    at full speed such as 350MB/sec (same speed as with non-thin lvm),\r\n"
- "    and also you don't see space occupation going up at every iteration\r\n"
- "    of dd, either with or without rm in-between the dd's. [ok actually\r\n"
- "    now retrying it needed 3 rewrites to stabilize allocation...\r\n"
- "    probably an AG count thing.]<br>\r\n"
- "    However the problem with XFS is that discard doesn't appear to work.\r\n"
- "    Fstrim doesn't work, and neither does \"mount -o discard ... + rm\r\n"
- "    zeroes\" . There is apparently no way to drop the allocated blocks,\r\n"
- "    as seen from lvs. This is in contrast to what it is written here\r\n"
- "    <meta http-equiv=\"content-type\" content=\"text/html; charset=UTF-8\">\r\n"
- "    <a href=\"http://xfs.org/index.php/FITRIM/discard\">http://xfs.org/index.php/FITRIM/discard</a>\r\n"
- "    which declare fstrim and mount -o discard to be working.<br>\r\n"
- "    Please note that since I am above MD raid5 (I believe this is the\r\n"
- "    reason), the passdown of discards does not work, as my dmesg says:<br>\r\n"
- "    [160508.497879] device-mapper: thin: Discard unsupported by data\r\n"
- "    device (dm-1): Disabling discard passdown.<br>\r\n"
- "    but AFAIU, unless there is a thinp bug, this should not affect the\r\n"
- "    unmapping of thin blocks by fstrimming xfs... and in fact ext4 is\r\n"
- "    able to do that.<br>\r\n"
- "    <br>\r\n"
- "    (*) Strange thing is that write performance appears to be roughly\r\n"
- "    the same for default thin chunksize and for 1MB thin chunksize. I\r\n"
- "    would have expected thinp allocation to be faster with larger thin\r\n"
- "    chunksizes but instead it is actually slower (note that there are no\r\n"
- "    snapshots here and hence no CoW). This is also true if I set the\r\n"
- "    thinpool to not zero newly allocated blocks: performances are about\r\n"
- "    240 MB/sec then, but again they don't increase with larger\r\n"
- "    chunksizes, they actually decrease slightly with very large\r\n"
- "    chunksizes such as 16MB. Why is that?<br>\r\n"
- "    <br>\r\n"
- "    Thanks for your help<br>\r\n"
- "    S.<br>\r\n"
- "  </body>\r\n"
- "</html>\r\n"
+ "<html>\n"
+ "  <head>\n"
+ "    <meta http-equiv=\"content-type\" content=\"text/html; charset=UTF-8\">\n"
+ "  </head>\n"
+ "  <body bgcolor=\"#FFFFFF\" text=\"#000000\">\n"
+ "    Hello all<br>\n"
+ "    I am doing some testing of dm-thin on kernel 3.4.2 and latest lvm\n"
+ "    from source (the rest is Ubuntu Precise 12.04).<br>\n"
+ "    There are a few problems with ext4 and (different ones with) xfs<br>\n"
+ "    <br>\n"
+ "    I am doing this:<br>\n"
+ "    dd if=/dev/zero of=zeroes bs=1M count=1000 conv=fsync<br>\n"
+ "    lvs<br>\n"
+ "    rm zeroes #optional<br>\n"
+ "    dd if=/dev/zero of=zeroes bs=1M count=1000 conv=fsync\302\240 #again<br>\n"
+ "    lvs<br>\n"
+ "    rm zeroes #optional<br>\n"
+ "    ...<br>\n"
+ "    dd if=/dev/zero of=zeroes bs=1M count=1000 conv=fsync\302\240 #again<br>\n"
+ "    lvs<br>\n"
+ "    rm zeroes<br>\n"
+ "    fstrim /mnt/mountpoint<br>\n"
+ "    lvs<br>\n"
+ "    <br>\n"
+ "    On ext4 the problem is that it always reallocates blocks at\n"
+ "    different places, so you can see from lvs that space occupation in\n"
+ "    the pool and thinlv increases at each iteration of dd, again and\n"
+ "    again, until it has allocated the whole thin device (really 100% of\n"
+ "    it). And this is true regardless of me doing rm or not between one\n"
+ "    dd and the other.<br>\n"
+ "    The other problem is that by doing this, ext4 always gets the worst\n"
+ "    performance from thinp, about 140MB/sec on my system, because it is\n"
+ "    constantly allocating blocks, instead of 350MB/sec which should have\n"
+ "    been with my system if it used already allocated regions (see below\n"
+ "    compared to xfs). I am on an MD raid-5 of 5 hdds.<br>\n"
+ "    I could suggest to add a \"thinp mode\" mount option to ext4 affecting\n"
+ "    the allocator, so that it tries to reallocate recently used and\n"
+ "    freed areas and not constantly new areas. Note that mount -o discard\n"
+ "    does work and prevents allocation bloating, but it still always gets\n"
+ "    the worst write performances from thinp. Alternatively thinp could\n"
+ "    be improved so that block allocation is fast :-P (*) <br>\n"
+ "    However, good news is that fstrim works correctly on ext4, and is\n"
+ "    able to drop all space allocated by all dd's. Also mount -o discard\n"
+ "    works.<br>\n"
+ "    <br>\n"
+ "    On xfs there is a different problem.<br>\n"
+ "    Xfs apparently correctly re-uses the same blocks so that after the\n"
+ "    first write at 140MB/sec, subsequent overwrites of the same file are\n"
+ "    at full speed such as 350MB/sec (same speed as with non-thin lvm),\n"
+ "    and also you don't see space occupation going up at every iteration\n"
+ "    of dd, either with or without rm in-between the dd's. [ok actually\n"
+ "    now retrying it needed 3 rewrites to stabilize allocation...\n"
+ "    probably an AG count thing.]<br>\n"
+ "    However the problem with XFS is that discard doesn't appear to work.\n"
+ "    Fstrim doesn't work, and neither does \"mount -o discard ... + rm\n"
+ "    zeroes\" . There is apparently no way to drop the allocated blocks,\n"
+ "    as seen from lvs. This is in contrast to what it is written here\n"
+ "    <meta http-equiv=\"content-type\" content=\"text/html; charset=UTF-8\">\n"
+ "    <a href=\"http://xfs.org/index.php/FITRIM/discard\">http://xfs.org/index.php/FITRIM/discard</a>\n"
+ "    which declare fstrim and mount -o discard to be working.<br>\n"
+ "    Please note that since I am above MD raid5 (I believe this is the\n"
+ "    reason), the passdown of discards does not work, as my dmesg says:<br>\n"
+ "    [160508.497879] device-mapper: thin: Discard unsupported by data\n"
+ "    device (dm-1): Disabling discard passdown.<br>\n"
+ "    but AFAIU, unless there is a thinp bug, this should not affect the\n"
+ "    unmapping of thin blocks by fstrimming xfs... and in fact ext4 is\n"
+ "    able to do that.<br>\n"
+ "    <br>\n"
+ "    (*) Strange thing is that write performance appears to be roughly\n"
+ "    the same for default thin chunksize and for 1MB thin chunksize. I\n"
+ "    would have expected thinp allocation to be faster with larger thin\n"
+ "    chunksizes but instead it is actually slower (note that there are no\n"
+ "    snapshots here and hence no CoW). This is also true if I set the\n"
+ "    thinpool to not zero newly allocated blocks: performances are about\n"
+ "    240 MB/sec then, but again they don't increase with larger\n"
+ "    chunksizes, they actually decrease slightly with very large\n"
+ "    chunksizes such as 16MB. Why is that?<br>\n"
+ "    <br>\n"
+ "    Thanks for your help<br>\n"
+ "    S.<br>\n"
+ "  </body>\n"
+ "</html>\n"
  "\01:2\0"
  "b\0"
  "_______________________________________________\n"
@@ -167,4 +167,4 @@
  "xfs@oss.sgi.com\n"
  http://oss.sgi.com/mailman/listinfo/xfs
 
-60ba8986651a8171009b26776bdabd6aec6dd8f51b4585316cb6ccff6843b805
+aaf5c244f8db0cbee0ff54a9d205d73f60f254d73a909cf10750f46314802101

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.