https://jira.whamcloud.com This file is an XML representation of an issue en-us 9.4.14 940014 05-12-2023 https://jira.whamcloud.com/browse/LU-6234 Lustre <p>When dealing with pools, you need to specify the file system name in front of the pool name as &lt;fsname&gt;.&lt;pool_name&gt;. The pool name, without the file system name, has a max length of 15 characters and is correctly enforced:</p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre># lctl pool_new scratch.test_82b Pool scratch.test_82b created # lctl pool_new scratch.test_82abcdefgh Pool scratch.test_82abcdefgh created # lctl pool_new scratch.test_82abcdefghi poolname test_82abcdefghi is too long (length is 16 max is 15) argument scratch.test_82abcdefghi must be &lt;fsname&gt;.&lt;poolname&gt; pool_new: File name too long </pre> </div></div> <p>Yet, when specifying a pool name in 'lfs setstripe', the file system name is used to compute the length of the pool name. The file system name is required when using the '-p' option in 'lfs setstripe' as in the second example below. </p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre># /usr/bin/lfs setstripe -p scratch.test_82b -o 7936 /lustre/scratch/d82b.conf-sanity/f82b.conf-sanity error: setstripe: pool name 'scratch.test_82b' is too long (max is 15 characters) ... # /usr/bin/lfs setstripe -p test_82b -o 4 /lustre/scratch/d82b.conf-sanity/f82b.conf-sanity can't find fs root for '/lustre/scratch/d82b.conf-sanity': -19 '/lustre/scratch/d82b.conf-sanity/f82b.conf-sanity' is not on a Lustre filesystem: No such device (19) error: setstripe: create stripe file '/lustre/scratch/d82b.conf-sanity/f82b.conf-sanity' failed ... </pre> </div></div> <p>This caused problems with the conf-sanity test 82b at <a href="https://testing.hpdd.intel.com/test_sets/e9152f7c-ade4-11e4-a0b6-5254006e85c2" class="external-link" target="_blank" rel="nofollow noopener">https://testing.hpdd.intel.com/test_sets/e9152f7c-ade4-11e4-a0b6-5254006e85c2</a> </p> lustre-master tag 2.6.93 LU-6234

lfs computes pool name length incorrectly

Bug Minor Resolved Fixed Zhenyu Xu James Nunez Wed, 11 Feb 2015 17:48:24 +0000 Thu, 16 Mar 2017 19:24:21 +0000 Tue, 18 Aug 2015 13:59:49 +0000 Lustre 2.7.0 Lustre 2.8.0 0 12 <p>This problem was introduced by change <a href="http://review.whamcloud.com/13241" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/13241</a> "<a href="https://jira.whamcloud.com/browse/LU-6081" title="hsm: add file migrate support" class="issue-link" data-issue-key="LU-6081">LU-6081</a> lfs: check that pool name is not too long". </p> <p>The problem is that the supplied pool name may include the FSNAME at the start, as is used in conf-sanity test_82b. In our test configuration we just happen to have the supplied pool name of "lustre.test_82b" which is shorter than the checked limit. The check in lfs_setstripe() should be changed to skip the leading fsname if present. </p> <p>That also begs the question - do we exclude '.' (and '/' and other problematic characters) from being part of the pool name?</p> <p>But does it make sense to pass a fsname? After all, the fsname is found through the file given in parameter.<br/> Should 6ec0b4 be reverted?</p> <p>It appears that specifying the pool name avoids the need to extract the fsname from the target directory, so it makes the code somewhat more efficient. Regardless, that is something that previously worked and was allowed, so it shouldn't be broken. It was only allowed to land because it happened that the <tt>fsname.pool_name</tt> was under the 15-char limit of the check. I'm OK with fixing the existing check to verify only the poolname part of the input is within the limits, or moving it lower in llapi to verify only the pool name after it has been stripped.</p> <p>Bobi Jam (bobijam@hotmail.com) uploaded a new patch: <a href="http://review.whamcloud.com/13742" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/13742</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-6234" title="lfs computes pool name length incorrectly" class="issue-link" data-issue-key="LU-6234"><del>LU-6234</del></a> lfs: check pool name length w/o fsname in it<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: a192f2660b86b3609b60bcc15911f4fd26d8ab7e</p> <p>Patch looks good. I've added a few tests to sanity.sh. See attachment. Would you mind folding them into that patch?</p> <p>A pool specific test script has been added (ost-pools.sh). New pool tests should go to it.</p> <p>Thanks Jean-Charles.</p> <p>I've updated the patch with tests in ost-pool.sh. See attachments.</p> <p>This is annoying. What is stalling the patch?</p> <p>Oleg Drokin (oleg.drokin@intel.com) merged in patch <a href="http://review.whamcloud.com/13742/" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/13742/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-6234" title="lfs computes pool name length incorrectly" class="issue-link" data-issue-key="LU-6234"><del>LU-6234</del></a> util: check fsname and pool name<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: 7c25eb1ba2b1db3009a0e88b3ecf229134f8ac92</p> <p>Landed for 2.8</p> <p>I noticed this patch landed, but it would have been good to add '.' to the characters that are not allowed in a pool or filesystem name. Otherwise it will break "lctl {get,set}&#95;param" using '.' as a separator.</p> <p>isalnum('.') == false, and current patch can handle it.</p> <p>You are right, thanks.</p> Related LU-6081 LU-7103 LU-7155 Development Rank 1|hzx66v: Rank (Obsolete) 17464 Severity