Whamcloud Community JIRA

Whamcloud Community JIRA https://jira.whamcloud.com This file is an XML representation of an issue en-us

9.4.14

940014

05-12-2023

[LU-11897] ineffective memory allocation in ptlrpc https://jira.whamcloud.com/browse/LU-11897 Lustre <p>For a few ptlrpc services, rqbd buffers are allocated using non-2^n allocation requests. This leads to ineffective memory usage and in some cases even to OOM.</p> <p>A patch will be uploaded shortly.</p> LU-11897

ineffective memory allocation in ptlrpc

Bug Major Resolved Fixed Andrew Perepechko Andrew Perepechko patch Tue, 29 Jan 2019 09:04:23 +0000 Tue, 19 Feb 2019 06:02:49 +0000 Tue, 19 Feb 2019 06:02:49 +0000 Lustre 2.13.0 0 2 <p>Andrew Perepechko (c17827@cray.com) uploaded a new patch: <a href="https://review.whamcloud.com/34127" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/34127</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-11897" title="ineffective memory allocation in ptlrpc" class="issue-link" data-issue-key="LU-11897"><del>LU-11897</del></a> ost: improve memory allocation for ost<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: c3c7734466d1b0b6ca248037e8eb4546371739f4</p> <p>How this bug led to an OOM event:</p> <blockquote> <p><span class="error">[12:19 PM]</span> Alexey Lyashkov: we have a 470k (+/-) rqbd buffers with 16k-32k size which hold just 660k rpc with 383 byte size.<br/> <span class="error">[12:20 PM]</span> Alexey Lyashkov: hm.. it looks we have a trivial bug in ptlrpc code<br/> <span class="error">[12:20 PM]</span> Alexey Lyashkov: #define OST_MAXREQSIZE (16 * 1024)<br/> <span class="error">[12:20 PM]</span> Alexey Lyashkov: but Panda say<br/> <span class="error">[12:21 PM]</span> Alexey Lyashkov: ```srv_max_req_size = 17408,```<br/> <span class="error">[12:21 PM]</span> Alexey Lyashkov: so MM uses a 32k allocation for 17k buffer<br/> <span class="error">[12:21 PM]</span> Alexey Lyashkov: and 15k is useless..<br/> <span class="error">[12:21 PM]</span> Alexey Lyashkov: and overhead..<br/> ...<br/> <span class="error">[12:42 PM]</span> Alexey Lyashkov: next catch is near of it - one more "17k" "leak"<br/> <span class="error">[12:42 PM]</span> Alexey Lyashkov: srv_buf_size = 17408, / srv_max_req_size = 17408,<br/> <span class="error">[12:43 PM]</span> Alexey Lyashkov: so 17k buffer will be unlinked after 328 bytes arrived.</p></blockquote> <p>While originally a whole 17 KiB (in fact, 32 KiB due to rounding) was used for a single 328 byte RPC, with the patch it is possible to put additional (32768-17408)/328-1=45 RPCs in the same buffer, no additional memory required.</p> <p>Oleg Drokin (green@whamcloud.com) merged in patch <a href="https://review.whamcloud.com/34127/" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/34127/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-11897" title="ineffective memory allocation in ptlrpc" class="issue-link" data-issue-key="LU-11897"><del>LU-11897</del></a> ost: improve memory allocation for ost<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: 3a90458bd84d43cc75c5a80f8c02f30d6412690a</p> <p>Landed for 2.13</p> Development Rank 1|i00ag7: Rank (Obsolete) 9223372036854775807 Severity