July 7, 2011 at 9:04 pm
Comments posted to this topic are about the item Self-aware Instances
July 7, 2011 at 10:11 pm
Hi Steve, I prefer your distributed approach as the workload is spread across the servers, and the data is discrete on each server. This approach provides a more granular backup and also simplifies rebuilding the aggregated data from its sources on the distributed servers. The central monitoring server then behaves like a data warehouse, gathering data from the distributed servers, performing extract-transform-load type tasks and managing the database. This architecture becomes more complex when you want to monitor activity in real time. However, you can retain the local copies of the data and also send a copy of the data to the monitoring server in real time.
July 8, 2011 at 2:34 am
Steve,
We use a home grown system (well several of them really) the main one polls each server hourly and pulls back all 'errors' (governed by a 'white/black' list) that have occurred in the last 24hr and send it as an email attachement to the dba group. The same job also brings back agent job errors. We have done it using a mixture of ssis, sql and .net.
Other home growns include distributed and centralised systems. So the information, data and code is distributed but then feeds in to a central repository that can be used for reporting (specific users then access this data via excel).
So there is deffinately a case for both techniques, though personally my option would be to have a solution which merges both.
We also have bought in solutions. Started with SQL Response but it wasnt great and we now use Idera DM, which I have to say SQL Monitor could learn a lot from.
Have a nice weekend, laters :hehe:
Adam Zacks-------------------------------------------Be Nice, Or Leave
July 8, 2011 at 8:24 am
Biology has a lot to teach us here.
Millipedes actually have pretty simple neural organization, each body segment has input from the segments in front and behind. So the 'brain' needs only to choose the direction and everything else follows along.
Ant, termite and bee colonies actually have fairly simple behavior patterns on the individual level. Each animal signals its immediate neighbors in a relatively limited language, but the colony as a whole displays very effective behavior built from the simple individuals.
It seems that many IT folks are psychologically 'top down' people (possibly because they were attracted to computers as the ultimate in controllable machines) and have a hard time 'letting go'
...
-- FORTRAN manual for Xerox Computers --
July 8, 2011 at 8:59 am
I like centralized. For me it makes maintenance easier. If I want to update a specific script, etc, I only have to change it in one place. Also DR is easier with a centralized solution. I'd rather have 1 DB holding the data instead of 1 on each instance. Less is more, IMO. Less parts. Less hassle. More efficiency. Although this is very generalized. I'm sure 'it depends' applies here as it does everywhere else.
-------------------------------------------------------------------------------------------------
My SQL Server Blog
July 8, 2011 at 9:44 am
In my mind it isn't a one OR the other proposition. If you have things that you want to monitor often it makes sense to have the instance gather and store that information for later pickup. For most things a daily pickup is fine. But I would expressly limit what I put on each instance because updates are much harder in this model since you have to touch each instance. I prefer the central command and control structure for most things. Especially for configuration management. If you are going to deploy things out to each instance I would strive to make the code as simple as possible so that the chances of changes to that code are minimized. The more complex it is the more likely you will have problems.
with all that said I use and like the Redgate SQL Monitor software. But it doesn't provide for a lot of the details I want and it makes sense for it not to. I want to capture and store many of the options about a database, options that have little performance impact, but I want to know when they change. Same for SQL Agent jobs, I want to know when these change and how. SQL Monitor does a great job of telling me they failed, or that their run duration was unusual but not that the job changed today.
And they build vs. buy is going to often end with Buy and build to fill in the voids not covered. The solution I have been working on doesn't provide real-time monitoring, it is mostly configuration management and logging. There is some overlap with SQL Monitor on database size tracking but much less granular..
CEWII
July 8, 2011 at 10:33 am
Elliott Whitlow (7/8/2011)
...If you have things that you want to monitor often it makes sense to have the instance gather and store that information for later pickup. For most things a daily pickup is fine. ...
I agree here. I prefer to store the info locally and then have something aggregate that data into a central location on a scheduled basis.
Jason...AKA CirqueDeSQLeil
_______________________________________________
I have given a name to my pain...MCM SQL Server, MVP
SQL RNNR
Posting Performance Based Questions - Gail Shaw[/url]
Learn Extended Events
July 8, 2011 at 10:42 am
SQLRNNR (7/8/2011)
Elliott Whitlow (7/8/2011)
...If you have things that you want to monitor often it makes sense to have the instance gather and store that information for later pickup. For most things a daily pickup is fine. ...I agree here. I prefer to store the info locally and then have something aggregate that data into a central location on a scheduled basis.
However, I always want to limit what I have to deploy out to the servers, so it is a balancing act.
CEWII
July 8, 2011 at 4:02 pm
Distributed.
July 9, 2011 at 9:29 am
Steve,
Got your message but cant reply as it says your over the limit 😉
Email me at work 🙂
Adam Zacks-------------------------------------------Be Nice, Or Leave
July 10, 2011 at 8:30 am
This will change based on the feedback and response time lag. Depending on priority, security, or ability to respond any system is best served by local correction. The more advanced local server software get the less corrections we will need to make from a external centralized controls.
Even God designed a dinosaurs with brains in its hip to control its legs because of neural speed. This answer is always changing...
July 11, 2011 at 12:00 am
It does not matter, all thou the later approach is interesting. The approach that takes the least time to maintain I'd go for.
Viewing 12 posts - 1 through 11 (of 11 total)
You must be logged in to reply to this topic. Login to reply