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-7796 Lustre <p><tt>lctl set_param jobid_var</tt> should return an error and print an error message about a missing value, but instead it is just ignored.</p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre>[root@centclient05 LU-7437-test]# lctl get_param jobid_var jobid_var=disable [root@centclient05 LU-7437-test]# lctl set_param jobid_var [root@centclient05 LU-7437-test]# echo $? 0 [root@centclient05 LU-7437-test]# lctl get_param jobid_var jobid_var=disable </pre> </div></div> <p>Note, however, that if the <tt>jobid_var</tt> has a trailing <tt>=</tt>, it behaves differently, despite meaning the same thing.</p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre>[root@centclient05 LU-7437-test]# lctl set_param jobid_var= error: set_param: setting jobid_var: no value [root@centclient05 LU-7437-test]# echo $? 22 </pre> </div></div> <p>Any trailing argument of the "set_param" without a corresponding value is ignored. E.g. the following variations also fail to print an error message when they should because jobid_var does not have a value:</p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre>lctl set_param ldlm.namespaces.*.lru_size=clear jobid_var lctl set_param ldlm.namespaces.*.lru_size clear jobid_var lctl set_param ldlm.namespaces.*osc*.lru_size=clear ldlm.namespaces.*mdc*.lru_size=clear jobid_var </pre> </div></div> <p>However, if jobid_var contains a trailing '=', all of these examples print an error message and return a non-zero status.</p> <p>In addition, this behavior is inconsistent and should be fixed:</p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre>[root@centclient05 LU-7437-test]# lctl set_param jobid_var=foo=bar jobid_var=foo=bar [root@centclient05 LU-7437-test]# lctl get_param jobid_var jobid_var=foo=bar [root@centclient05 LU-7437-test]# lctl set_param jobid_var foo=bar error: set_param: setting jobid_var=foo=bar: bad value [root@centclient05 LU-7437-test]# echo $? 22 [root@centclient05 LU-7437-test]# lctl get_param jobid_var jobid_var=foo=bar </pre> </div></div> <p>The two different <tt>lctl set_param</tt> commands above should mean the same thing. IMO, lctl set_param should accept values which contain the <tt>=</tt> character because there is no intrinsic constraint against writing a <tt>=</tt> to the corresponding procfs/sysfs/debugfs file. </p> LU-7796

"lctl set_param jobid_var" should return error about missing value

Bug Minor Resolved Fixed Emoly Liu Ryan Haasken Thu, 18 Feb 2016 21:51:10 +0000 Mon, 11 Apr 2016 04:39:25 +0000 Mon, 11 Apr 2016 04:39:25 +0000 Lustre 2.9.0 0 6 <p>This was introduced by this patch: <a href="http://review.whamcloud.com/#/c/18338/12" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#/c/18338/12</a></p> <p>I am willing to fix this, and I will provide a patch shortly.</p> <p>Please let me know if you disagree with what I believe to be the correct behavior for lctl set_param. Namely, that it should print an error message and return a nonzero status for trailing valueless parameters of the space-separated form, and that it should allow for <tt>=</tt> characters in the values in the space-separated format.</p> <p>Ryan, when we use "lctl set_param" to set one parameter in one command, we can use the following two formats:</p> <ul class="alternate" type="square"> <li>lctl set_param param_path1=value1</li> <li>or old format, lctl set_param param_path1 value1</li> </ul> <p>And when we set multiple parameters, we recommend to use the following format</p> <ul class="alternate" type="square"> <li>lctl set_param &lt;param_path1=value1 param_path2=value2 ...&gt;. However, space-separated format is still supported in this case, but this format is not easy to read.</li> </ul> <p>IMO, "=" in the values will get the format mixed up. <a href="https://jira.whamcloud.com/secure/ViewProfile.jspa?name=adilger" class="user-hover" rel="adilger">adilger</a>, do you have advice on this format?</p> <p>BTW, "lctl set_param param_path" should return error since there is no valid value. Thanks for pointing this out.</p> <p>My preference is that we don't allow set_param with multiple space-separated arguments, and only as a backward-compatible way of handling single "name value" arguments for older scripts that do not use "name=value". As Emoly writes, the chance of having errors with "name1 value1 name2 value2 name3 value3" becomes significant, doubly so if the values allow an embedded '=' in them (which I had never intended, but I guess was never explicitly forbidden).</p> <p>I mentioned in another patch that I think "name value" arguments should be deprecated in 2.9 (i.e print a warning to stderr that they should use "name=value") and this format should be dropped completely at some point in the future (say 3.2 or whatever).</p> <p>Thank you both for your input. I agree that the format with parameters and values separated by an '=' is better than the format with them separated by a space. I still plan to implement the deprecation warning for the old format in the patch for <a href="https://jira.whamcloud.com/browse/LU-5134" title="Add option to lctl set_param for setting parameters in parallel" class="issue-link" data-issue-key="LU-5134"><del>LU-5134</del></a>.</p> <blockquote> <p>My preference is that we don't allow set_param with multiple space-separated arguments, and only as a backward-compatible way of handling single "name value" arguments for older scripts that do not use "name=value".</p></blockquote> <p>I don't think it would be acceptable to change the implementation to reject the old format if more than one parameter-value pair is specified. This would be a change in the actual behavior that could break people's scripts, regardless of the recommended format for multiple param-value pairs.</p> <p>I also see no reason why an '=' should be explicitly forbidden from appearing in the value. It is possible (although probably not currently required by any Lustre/LNet params) that a value contains an '='. I think the current behavior of an <tt>lctl set_param param_path1=value1</tt> when <tt>value1</tt> contains an '=' character is correct, and the older format should behave the same way. This is a minor issue. The bigger issue is probably that <tt>lctl set_param param_path1</tt> does not print an error message about a missing value and exit with a nonzero status.</p> <p>Ryan, I agree to fix that "bigger" issue first. Would you like to provide a patch for that? Or I can do that.</p> <p>I can provide such a patch. Would it make sense to also add the deprecation warning in this patch instead of in the patch for <a href="https://jira.whamcloud.com/browse/LU-5134" title="Add option to lctl set_param for setting parameters in parallel" class="issue-link" data-issue-key="LU-5134"><del>LU-5134</del></a>? This would only be desirable if this fix will NOT land for 2.8. See line 1725 here for context:</p> <p><a href="http://review.whamcloud.com/#/c/10555/5/lustre/utils/lustre_cfg.c" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#/c/10555/5/lustre/utils/lustre_cfg.c</a></p> <p>Then see PS6 of that change for a possible implementation of the deprecation warning. I am undecided on whether it still makes sense to separate the parsing of the param/value pair into a new function in the fix for this bug. I think it's good to keep the functions shorter, but one could argue it creates an unnecessarily complicated interface in the code. I'll upload a possible patch and look for your feedback this week.</p> <p>Emoly Liu (emoly.liu@intel.com) uploaded a new patch: <a href="http://review.whamcloud.com/18684" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/18684</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-7796" title="&quot;lctl set_param jobid_var&quot; should return error about missing value" class="issue-link" data-issue-key="LU-7796"><del>LU-7796</del></a> lctl: set_param should report errors for wrong formats<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: dce7003107212cfb07eaa9e968d27a2dfaad9557</p> <p>Ryan, I know you are working on <a href="https://jira.whamcloud.com/browse/LU-5134" title="Add option to lctl set_param for setting parameters in parallel" class="issue-link" data-issue-key="LU-5134"><del>LU-5134</del></a> <a href="http://review.whamcloud.com/#/c/10555/" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#/c/10555/</a>. In case that your big patch won't be landed soon to fix this issue, I created this one. Please review it.</p> <blockquote> <p><a href="https://jira.whamcloud.com/browse/LU-7796" title="&quot;lctl set_param jobid_var&quot; should return error about missing value" class="issue-link" data-issue-key="LU-7796"><del>LU-7796</del></a> lctl: set_param should report errors for wrong formats</p> <p>"lctl set_param" should report errors for the following wrong<br/> parameter formats:</p> <ul class="alternate" type="square"> <li>path</li> <li>path=</li> <li>mix of the above formats, e.g.<br/> "lctl set_param jobid_var= timeout=20 abde=efg debug"</li> </ul> </blockquote> <p>Thanks, Emoly. I had actually planned to submit the fix for this issue in a separate patch from my patch for <a href="https://jira.whamcloud.com/browse/LU-5134" title="Add option to lctl set_param for setting parameters in parallel" class="issue-link" data-issue-key="LU-5134"><del>LU-5134</del></a>. I reviewed your patch. There are a couple of things I don't agree with about the patch you've uploaded, particularly the handling of equals signs in the values when using the older <tt>set_param param value</tt> format. I will upload my patch too, and if you could take a look, I would appreciate it. I would prefer if we could land my patch since I think it will be a little cleaner to add the deprecation of the old <tt>set_param param value</tt> format to my patch. I also had already written up some <tt>sanity.sh</tt> tests for the problems listed in this bug, which I will include in my patch.</p> <p>Ryan Haasken (haasken@cray.com) uploaded a new patch: <a href="http://review.whamcloud.com/18703" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/18703</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-7796" title="&quot;lctl set_param jobid_var&quot; should return error about missing value" class="issue-link" data-issue-key="LU-7796"><del>LU-7796</del></a> utils: print error with valueless lctl set_param<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: 2f86b817ad4d73c4f2c6112d743d770b8250898a</p> <p>I pushed my version of a fix for this bug here:</p> <p><a href="http://review.whamcloud.com/#/c/18703" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#/c/18703</a></p> <p>Please take a look. I think it simplifies the code a little. I tried to compromise on allowing '=' in the values in the 'set_param param value' format by printing a warning message in that case.</p> <p>Oleg Drokin (oleg.drokin@intel.com) merged in patch <a href="http://review.whamcloud.com/18703/" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/18703/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-7796" title="&quot;lctl set_param jobid_var&quot; should return error about missing value" class="issue-link" data-issue-key="LU-7796"><del>LU-7796</del></a> utils: print error with valueless lctl set_param<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: e2973925ca550ad8ef7ad9af72d7d8766f9c1f41</p> <p>Landed for 2.9.</p> Related LU-7437 LU-7802 Development Rank 1|hzy1tj: Rank (Obsolete) 9223372036854775807 Severity