<div dir="ltr">Good news, after upgrading to 5.13 and running this scenario again, the self heal actually succeeded without my intervention following a server reboot.<div><br></div><div>The load was still high during this process, but at least the endless heal issue is resolved.</div><div><br></div><div>I&#39;d still love to hear from the team on managing heal load spikes.<br clear="all"><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><br>Sincerely,<br>Artem<br><br>--<br>Founder, <a href="http://www.androidpolice.com" target="_blank">Android Police</a>, <a href="http://www.apkmirror.com/" style="font-size:12.8px" target="_blank">APK Mirror</a><span style="font-size:12.8px">, Illogical Robot LLC</span></div><div dir="ltr"><a href="http://beerpla.net/" target="_blank">beerpla.net</a> | <a href="http://twitter.com/ArtemR" target="_blank">@ArtemR</a><br></div></div></div></div></div></div></div></div></div></div></div></div></div><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Apr 26, 2020 at 3:13 PM Artem Russakovskii &lt;<a href="mailto:archon810@gmail.com">archon810@gmail.com</a>&gt; wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div>Hi all,<br></div><div><br></div><div>I&#39;ve been observing this problem for a long time now and it&#39;s time to finally figure out what&#39;s going on.</div><div><br></div><div>We&#39;re running gluster 5.11 and have a 10TB 1 x 4 = 4 replicate volume. I&#39;ll include its slightly redacted config below.</div><div><br></div><div>When I reboot one of the servers and it goes offline for a bit, when it comes back, heal info tells me there are some files and dirs that are &quot;heal pending&quot;. 0 &quot;split-brain&quot; and &quot;possibly healing&quot; - only &quot;heal pending&quot; are &gt;0.</div><div><ol><li>For some reason, the server that was rebooted shows &quot;heal pending&quot; 0. All other servers show &quot;heal pending&quot; with some number, say 65.</li><li>We have cluster.self-heal-daemon enabled.</li><li>The logs are full of &quot;performing entry selfheal&quot; and &quot;completed entry selfheal&quot; messages that continue to print endlessly.</li><li>This &quot;heal pending&quot; number never goes down by itself, but it does if I run some operation on it, like md5sum.</li><li>When the server goes down for reboot and especially when it comes back, the load on ALL servers shoots up through the roof (load of 100+) and ends up bringing everything down, including apache and nginx. My theory is that self-heal kicks in so hard that it kills IO on these attached Linode block devices. However, after some time - say 10 minutes - the load subsides, but the &quot;heal pending&quot; remains and the gluster logs continue to output

&quot;performing entry selfheal&quot; and &quot;completed entry selfheal&quot; messages. This load spike has become a huge issue for us because it brings down the whole site for entire minutes.</li><li>At this point in my investigation, I noticed that the selfheal messages actually repeat for the same gfids over and over.<br><font face="monospace">[2020-04-26 21:32:29.877987] I [MSGID: 108026] [afr-self-heal-entry.c:897:afr_selfheal_entry_do] 0-SNIP_data1-replicate-0: performing entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc<br>[2020-04-26 21:32:29.901246] I [MSGID: 108026] [afr-self-heal-common.c:1729:afr_log_selfheal] 0-SNIP_data1-replicate-0: Completed entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc. sources= sinks=0 1 2<br>[2020-04-26 21:32:32.171959] I [MSGID: 108026] [afr-self-heal-entry.c:897:afr_selfheal_entry_do] 0-SNIP_data1-replicate-0: performing entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc<br>[2020-04-26 21:32:32.225828] I [MSGID: 108026] [afr-self-heal-common.c:1729:afr_log_selfheal] 0-SNIP_data1-replicate-0: Completed entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc. sources= sinks=0 1 2<br>[2020-04-26 21:32:33.346990] I [MSGID: 108026] [afr-self-heal-entry.c:897:afr_selfheal_entry_do] 0-SNIP_data1-replicate-0: performing entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc<br>[2020-04-26 21:32:33.374413] I [MSGID: 108026] [afr-self-heal-common.c:1729:afr_log_selfheal] 0-SNIP_data1-replicate-0: Completed entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc. sources= sinks=0 1 2</font></li><li>I used gfid-resolver.sh from <a href="https://gist.github.com/4392640.git" target="_blank">https://gist.github.com/4392640.git</a> to resolve this gfid to the real location and yup - it was one of the files (a dir actually) listed as &quot;heal pending&quot; in heal info. As soon as I ran md5sum on the file inside (which was also listed in &quot;heal pending&quot;), the log messages stopped repeating for this entry and it disappeared from &quot;heal pending&quot; heal info. These were the final log lines:<br><font face="monospace">[2020-04-26 21:32:35.642662] I [MSGID: 108026] [afr-self-heal-metadata.c:52:__afr_selfheal_metadata_do] 0-SNIP_data1-replicate-0: performing metadata selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc<br>[2020-04-26 21:32:35.658714] I [MSGID: 108026] [afr-self-heal-common.c:1729:afr_log_selfheal] 0-SNIP_data1-replicate-0: Completed metadata selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc. sources=0 [1] 2  sinks=3<br>[2020-04-26 21:32:35.686509] I [MSGID: 108026] [afr-self-heal-entry.c:897:afr_selfheal_entry_do] 0-SNIP_data1-replicate-0: performing entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc<br>[2020-04-26 21:32:35.720387] I [MSGID: 108026] [afr-self-heal-common.c:1729:afr_log_selfheal] 0-SNIP_data1-replicate-0: Completed entry selfheal on 96d282cf-402f-455c-9add-5f03c088a1bc. sources=0 [1] 2  sinks=3</font></li></ol></div><div>I have to repeat this song and dance every time I reboot servers and run md5sum on each &quot;heal pending&quot; file or else the messages will continue presumably indefinitely. In the meantime, the files seem to be fine when accessed.</div><div><br></div><div>What I don&#39;t understand is:</div><div><ol><li>Why doesn&#39;t 

gluster 

just heal them properly instead of getting stuck? Or maybe this was fixed in v6 or v7, which I haven&#39;t upgraded to due to waiting for another unrelated issue to be fixed?</li><li>Why does heal info show 0 &quot;heal pending&quot; files on the server that was rebooted, but all other servers show the same number of &quot;heal pending&quot; entries &gt;0?</li><li>Why are there these insane load spikes upon going down and especially coming back online? Is it related to the issue here? I&#39;m pretty sure that it didn&#39;t happen in previous versions of gluster, when this issue didn&#39;t manifest - I could easily bring down one of the servers without it creating havoc when it comes back online.</li></ol></div><div>Here&#39;s the volume info:<br></div><blockquote style="margin:0px 0px 0px 40px;border:none;padding:0px"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Volume Name: SNIP_data1</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Type: Replicate</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Volume ID: 11ecee7e-d4f8-497a-9994-ceb144d6841e</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Status: Started</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Snapshot Count: 0</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Number of Bricks: 1 x 4 = 4</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Transport-type: tcp</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Bricks:</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Brick1: SNIP:/mnt/SNIP_block1/SNIP_data1</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Brick2: SNIP:/mnt/SNIP_block1/SNIP_data1</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Brick3: SNIP:/mnt/SNIP_block1/SNIP_data1</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Brick4: SNIP:/mnt/SNIP_block1/SNIP_data1</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Options Reconfigured:</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.client-io-threads: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">nfs.disable: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">transport.address-family: inet</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.self-heal-daemon: enable</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.cache-size: 1GB</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.lookup-optimize: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.read-ahead: off</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">client.event-threads: 4</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">server.event-threads: 4</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.io-thread-count: 32</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.readdir-optimize: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">features.cache-invalidation: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">features.cache-invalidation-timeout: 600</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.stat-prefetch: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.cache-invalidation: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.md-cache-timeout: 600</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">network.inode-lru-limit: 500000</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.parallel-readdir: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.readdir-ahead: on</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">performance.rda-cache-limit: 256MB</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">network.remote-dio: enable</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">network.ping-timeout: 5</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.quorum-type: fixed</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.quorum-count: 1</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.granular-entry-heal: enable</blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">cluster.data-self-heal-algorithm: full</blockquote></blockquote><div><br></div><div>Appreciate any insight. Thank you.</div><div><div><div dir="ltr"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><br>Sincerely,<br>Artem<br><br>--<br>Founder, <a href="http://www.androidpolice.com" target="_blank">Android Police</a>, <a href="http://www.apkmirror.com/" style="font-size:12.8px" target="_blank">APK Mirror</a><span style="font-size:12.8px">, Illogical Robot LLC</span></div><div dir="ltr"><a href="http://beerpla.net/" target="_blank">beerpla.net</a> | <a href="http://twitter.com/ArtemR" target="_blank">@ArtemR</a><br></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div>
</blockquote></div>