Attachment 299362 Details for Bug 220060 – Documentation adjustment to describe cull limitations

[patch] Documentation adjustment to describe cull limitations

cachefilesd-0.8-cull-doc.patch (text/plain), 3.97 KB, created by David Howells on 2008-03-27 17:14:43 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: David Howells

Created: 2008-03-27 17:14:43 UTC

Size: 3.97 KB

patch

obsolete

>diff -ur cachefilesd-0.8-orig/cachefilesd.conf.5 cachefilesd-0.8/cachefilesd.conf.5
>--- cachefilesd-0.8-orig/cachefilesd.conf.5	2006-11-14 18:04:09.000000000 +0000
>+++ cachefilesd-0.8/cachefilesd.conf.5	2008-03-27 17:03:13.000000000 +0000
>@@ -106,9 +106,20 @@
> .SH CACHE CULLING
> .P
> The cache may need culling occasionally to make space.  This involves
>-discarding objects from the cache that have been used less recently than
>-anything else.  Culling is based on the access time of data objects.  Empty
>-directories are culled if not in use.
>+discarding objects from the cache that are no longer marked as being in use by
>+the network filesystem.  Objects are selected from this group based on which
>+has the least recent access time.  Empty directories are culled if not in use.
>+.P
>+Note that this culling policy may result in objects remaining in cache longer
>+than one might expect, because all references to the file by the network
>+filesystem must be released before the file is culled.  In the case of NFS,
>+this means that the in-memory record of the inode must be discarded.  This
>+record will be pinned in memory whilst the inode has open file descriptors on
>+it or bits of the inode are memory mapped.
>+.P
>+Once all references to it are released, the record will be retained in memory
>+until memory pressure causes it to be discarded, or the inode is deleted or the
>+filesystem is unmounted.
> .P
> Cache culling is done on the basis of the percentage of blocks and the
> percentage of files available in the underlying filesystem.  There are six
>diff -ur cachefilesd-0.8-orig/howto.txt cachefilesd-0.8/howto.txt
>--- cachefilesd-0.8-orig/howto.txt	2006-11-13 14:52:59.000000000 +0000
>+++ cachefilesd-0.8/howto.txt	2008-03-27 17:08:59.000000000 +0000
>@@ -205,6 +205,14 @@
> cache until appropriate parameter falls back beneath the run limit.
> 
> 
>+Note that culling is limited by the retention of in-memory inode records by the
>+Linux VFS layer.  Even if an inode is not currently being used, the VFS may
>+retain a record of it to avoid having to reget the inode later.  If such an
>+inode is pinning an object in the cache, then that object may not be culled
>+until the VFS asks the network filesystem to discard the inode for whatever
>+reason (deletion, server synchronisation, unmounting or memory pressure).
>+
>+
> ==========
> MONITORING
> ==========
>diff -ur cachefilesd-0.8-orig/README cachefilesd-0.8/README
>--- cachefilesd-0.8-orig/README	2006-11-14 17:47:37.000000000 +0000
>+++ cachefilesd-0.8/README	2008-03-27 17:11:22.000000000 +0000
>@@ -16,6 +16,8 @@
> 
>  (*) Cache culling.
> 
>+ (*) Cull limitations.
>+
>  (*) Cache structure.
> 
>  (*) Security model and SELinux.
>@@ -46,10 +48,10 @@
> 
> CacheFiles attempts to maintain at least a certain percentage of free space on
> the filesystem, shrinking the cache by culling the objects it contains to make
>-space if necessary - see the "Cache Culling" section.  This means it can be
>-placed on the same medium as a live set of data, and will expand to make use of
>-spare space and automatically contract when the set of data requires more
>-space.
>+space if necessary - see the "Cache Culling" and "Cull limitations" sections.
>+This means it can be placed on the same medium as a live set of data, and will
>+expand to make use of spare space and automatically contract when the set of
>+data requires more space.
> 
> 
> ============
>@@ -223,6 +225,18 @@
> their atimes have changed or if the kernel module says it is still using them.
> 
> 
>+================
>+CULL LIMITATIONS
>+================
>+
>+Note that culling is limited by the retention of in-memory inode records by the
>+Linux VFS layer.  Even if an inode is not currently being used, the VFS may
>+retain a record of it to avoid having to reget the inode later.  If such an
>+inode is pinning an object in the cache, then that object may not be culled
>+until the VFS asks the network filesystem to discard the inode for whatever
>+reason (deletion, server synchronisation, unmounting or memory pressure).
>+
>+
> ===============
> CACHE STRUCTURE
> ===============

Actions: View | Diff

Attachments on bug 220060: 299362