<div class="gmail_quote">On Fri, Mar 18, 2011 at 2:28 PM, Matthew Steven <span dir="ltr"><<a href="mailto:matthew@geniusweb.com">matthew@geniusweb.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On 03/18/2011 12:48 PM, Matthew Nuzum wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Matt, you take the snapshots outside of the VM? Doesn't that cause potential for consistency problems?<br>
<br>
</blockquote></div>
Absolutely.<br>
<br>
Even though LVM sync's everything to disk on the original volume before the snapshot COW starts, that doesn't necessarily guarantee consistency since the database will do with the data as it wilt.<br>
<br>
Coordinating a flush tables/lock before the snapshot creation is a good way to handle that. Or if that's likely to be mis-timed or inconvenient in your deployment, you can also just ensure that the system dumps the tables to disk regularly. Or use the db's built in syncronization. The list goes on and on!<div>
<div></div><div class="h5"><br></div></div></blockquote></div><br>You say that so candidly. Let me just confirm that what I understand from your statement is that using this technique requires caution and probably bears the same hassles as rebooting a machine that had the power cord pulled, right?<br clear="all">
<br>-- <br>Matthew Nuzum<br>newz2000 on freenode, skype, linkedin, <a href="http://identi.ca" target="_blank">identi.ca</a> and twitter<br><br>"An investment in knowledge pays the best interest." -Benjamin Franklin <br>
<br>