RE: Write cache for a SAN
Date: Tue, 4 Nov 2008 00:13:34 -0500
So, having been on the vendor end of many large customer implementations while I was at EMC, and seeing a couple of horrifying customer support situations, here's my opinions on the subject:
- hardware fails like software - inexplicably, and at the least convenient time
- the people who get sent out to fix the problem will be commensurate with the corporate perception of the difficulty of the problem
- people always make mistakes - every vendor will, and no one can tell you they won't
- customers often explain the problem they think they're having without sufficient detail
Hence, I often recommend:
- have engineering-level contacts at your vendors (before I worked at EMC and was an EMC customer, I escalated to an engineering level to hear them say, 'What the f*** are you trying to do? Why would you try to do that?" - (because I had happened across an EMC-only config, and she assumed I was an employee)
- If a field engineer arrives onsite and you don't trust them, send them back. Think of it like fishing - if they're not a legal size, you throw them back
- know more than they do - quiz support folks and field people - if you don't trust them, find people who do and get them to help you.
I recommend this methodology for not just storage, but everything tech related. Make valuable relationships, and keep thm in your back pocket. I have customers who call me with Linux issues randomly because they know I'm not going to make stuff up and will give an honest answer.
From: oracle-l-bounce_at_freelists.org on behalf of Jared Still Sent: Mon 11/3/2008 3:49 PM
To: Bobak, Mark
Cc: dofreeman_at_state.pa.us; Oracle-L Freelists Subject: Re: Write cache for a SAN
On Mon, Nov 3, 2008 at 12:29 PM, Bobak, Mark <Mark.Bobak_at_proquest.com> wrote:
This really is the exception, though, and not the rule. I don't think disabling write caching is the right move. Ultimately, at some point, you have to have some faith in the hardware you've got and the people you work with.
I personally know of several instances of such things that make this type of thing much less an exception, and more of a rule.
You may have seen that I did not recommend disabling write cache.
My recommendation is to make sure the tech on site knows what he is doing.
How do you do that?
Get the same instructions he is supposed to be following, and make sure that he does indeed follow them.
I don't doubt the technology works.
The problem is often the nut on the other end of the wrench.
Certifiable Oracle DBA and Part Time Perl Evangelist