<div dir="ltr">Hi Digimer<div><br></div><div>I found in the test that there will be a 10 second heartbeat wait after the </div><div>drbd synchronization network port is down. During this time, the new </div><div>data IO will not be synchronized to the other end. When 10 seconds later, </div><div>these request that waiting for synchronization include the master BIO </div><div>will be cleaned up and then returned correctly when the clone BIO </div><div>corresponding to these master BIOs is written to disk. In this case, </div><div>will the pacemaker with a stonith configuration finally decide to stop </div><div>the controller? But another controller does not have this data, which </div><div>leads to data loss.</div></div><br><div class="gmail_quote"><div dir="ltr">Digimer <<a href="mailto:lists@alteeve.ca">lists@alteeve.ca</a>> 于2018年12月10日周一 下午3:30写道：<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 2018-12-09 8:28 p.m., Su Hua wrote:<br>
> Excuse me, the latest drbd driver can provide the arbitration service in<br>
> the case of split-brain？instead of using the script that is fixed<br>
> afterwards, which may cause partial data loss under the GFS2 cluster<br>
> file system.<br>
<br>
DRBD needs to be fenced in a way that informs DLM that the lost node has<br>
been fenced. I do not know of any fence handlers that do this is a pure<br>
DRBD install. So for practical purposes, no, you will need pacemaker<br>
with a proper stonith configuration to avoid split-brains in the first<br>
place.<br>
<br>
-- <br>
Digimer<br>
Papers and Projects: <a href="https://alteeve.com/w/" rel="noreferrer" target="_blank">https://alteeve.com/w/</a><br>
"I am, somehow, less interested in the weight and convolutions of<br>
Einstein’s brain than in the near certainty that people of equal talent<br>
have lived and died in cotton fields and sweatshops." - Stephen Jay Gould<br>
</blockquote></div>