1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
|
<?xml version='1.0' encoding='UTF-8'?>
<!-- This document was created with Syntext Serna Free. --><!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN" "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" [
<!ENTITY % BOOK_ENTITIES SYSTEM "Release_Notes.ent">
%BOOK_ENTITIES;
]>
<chapter id="chap-Release_Notes-Known_Issues">
<title>Known Issues</title>
<para>
The following are the known issues:
</para>
<para><itemizedlist>
<listitem>
<para>Issues related to Distributed Replicated Volumes:
</para>
<itemizedlist>
<listitem>
<para>When process has done <emphasis role="italic">
<emphasis role="italic">
<command>cd</command>
</emphasis>
</emphasis> into a directory, stat of deleted file recreates it (directory self-
heal not triggered).
</para>
<para>In GlusterFS replicated setup, if you are inside a directory (for example, <filename>Test</filename> directory) of
replicated volume. From another node, you will delete a file inside <filename>Test</filename> directory. Then if you
perform <command>stat</command> operation on the same file name, the file will be automatically created. (that is, a
proper directory self-heal is not triggered when process has done <command>cd</command> into a path).
</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>Issues related to Distributed Volumes:
</para>
<itemizedlist>
<listitem>
<para>Rebalance does not happen if bricks are down.
</para>
<para>Currently while running rebalance, make sure all the bricks are in operating or connected state.
</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>glusterfsd - Error return code is not proper after daemonizing the process.
</para>
<para>Due to this, scripts that mount glusterfs or start glusterfs process must not depend on its return
value.
</para>
</listitem>
<listitem>
<para>After <command># gluster volume replace-brick <replaceable>VOLNAME Brick New-Brick</replaceable> commit</command> command
is issued, the file system operations on that particular volume, which are in transit will fail.
</para>
</listitem>
<listitem>
<para>Command <command># gluster volume replace-brick ...</command> will fail in a RDMA set up.
</para>
</listitem>
<listitem>
<para>If files and directories have different GFIDs on different backends, GlusterFS client may hang or
display errors.
</para>
<para><emphasis role="bold">Work Around</emphasis>: The workaround for this issue is explained at <ulink url="http://gluster.org/pipermail/gluster-users/2011-July/008215.html"/>
.
</para>
</listitem>
<listitem>
<para>Issues related to Directory Quota:
</para>
<itemizedlist>
<listitem>
<para>Some writes can appear to pass even though the quota limit is exceeded (write returns
success). This is because they could be cached in write-behind. However disk-space would
not exceed the quota limit, since when writes to backend happen, quota does not allow
them. Hence it is advised that applications should check for return value of close call.
</para>
</listitem>
<listitem>
<para>If a user has done <command>cd</command> into a directory on which the administrator is setting the limit, even
though the command succeeds and the new limit value will be applicable to all the users
except for those users’ who has done <command>cd</command> in to that particular directory. The old limit value
will be applicable until the user has <command>cd</command> out of that directory.
</para>
</listitem>
<listitem>
<para>Rename operation (that is, removing oldpath and creating newpath) requires additional disk
space equal to file size. This is because, during rename, it subtracts the size on oldpath after
rename operation is performed, but it checks whether quota limit is exceeded on parents of
newfile before rename operation.
</para>
</listitem>
<listitem>
<para>With striped volumes, Quota feature is not available.</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>Issues related to POSIX ACLs:
</para>
<itemizedlist>
<listitem>
<para>Even though POSIX ACLs are set on the file or directory, the <computeroutput>+</computeroutput> (plus) sign in the file
permissions will not be displayed. This is for performance optimization and will be fixed in a
future release.
</para>
</listitem>
</itemizedlist>
<itemizedlist>
<listitem>
<para>When glusterfs is mounted with <command>-o acl</command>, directory read performance can be bad. Commands
like recursive directory listing can be slower than normal.
</para>
</listitem>
</itemizedlist>
<itemizedlist>
<listitem>
<para>When POSIX ACLs are set and multiple NFS clients are used, there could be inconsistency in
the way ACLs are applied due to attribute caching in NFS. For a consistent view of POSIX ACLs
in a multiple client setup, use <command>-o noac</command> option on NFS mount to switch off attribute caching.
This could have a performance impact on operations involving attributes.
</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>If you have enabled Gluster NLM, you cannot mount kernel NFS client on your storage nodes.
</para>
</listitem>
<listitem condition="gfs">
<para>Error with lost and found directory while using multiple disks.</para>
<para><emphasis role="bold">Work Around</emphasis>: You must ensure that the brick directories are one of the extra directories in the back-end mount point. For Example, if <filename>/dev/sda1</filename> is mounted on <filename>/export1</filename>, use <filename>/export1/volume</filename> as the glusterfs's export directory. </para>
</listitem>
<listitem>
<para>Due to enhancements in Graphs, you may experience excessive memory usage with this release.
</para>
</listitem>
<listitem>
<para>After you restart the NFS server, the unlock within the grace-period may fail and previously held locks may not be reclaimed.
</para>
</listitem>
<listitem>
<para>After a rebalancing a volume, if you run <command>rm -rf</command> command at the mount point to remove all contents of the current working directory recursively without prompting, you may get "Directory not Empty" error message.
</para>
</listitem>
<listitem>
<para>The following is a known missing (minor) feature:
</para>
<itemizedlist>
<listitem>
<para>locks - <emphasis role="italic">mandatory</emphasis> locking is not supported.
</para>
</listitem>
</itemizedlist>
</listitem>
</itemizedlist></para>
</chapter>
|
