Timeline for Page Life Expectancy (PLE), where to start?
Current License: CC BY-SA 3.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 30, 2019 at 15:01 | answer | added | James Jenkins | timeline score: 1 | |
| Apr 13, 2018 at 9:36 | history | tweeted | twitter.com/StackDBAs/status/984726897450942466 | ||
| Apr 13, 2018 at 9:09 | history | edited | James Jenkins | CC BY-SA 3.0 |
added 420 characters in body
|
| Apr 12, 2018 at 18:08 | comment | added | Aaron Bertrand | Well, you can just ignore PLE. You are not observing a problem, your users are not complaining about a problem, so don't let some PLE number make you chase unicorns. Make SCOM alert you on actual symptoms, like high waits in a specific category or severe drops in actual buffer pool usage, not an imaginary 300 threshold of PLE on its own. What you need to monitor for specifically will depend a lot more on your workload, your system, your usage patterns, etc. than on some blind list your peers might provide to you. | |
| Apr 12, 2018 at 18:04 | comment | added | James Jenkins | @AaronBertrand I concur, I am with you 100% on that last comment. I am looking at all 4 of the vitals in combination, and I am still pretty sure I don't have an issue (on the instance that started me on this quest). It's a VM I could just throw some Memory at it and I would probably stop getting the PLE alerts. But I don't want to just do that. I have checked waits, I have checked indexes, I balance the existing memory better (with OS), I really think it is fine. I know I don't want to look at PLE alone, but I am not sure what else (related) to look at with it. | |
| Apr 12, 2018 at 17:45 | comment | added | Aaron Bertrand | I guess what I'm suggesting is that you don't need to panic every time PLE drops. There isn't always going to be a problem to solve. I wouldn't have anything set up to alert on PLE alone because it is so rarely an indication of some problem I can solve, or a problem at all. | |
| Apr 12, 2018 at 17:35 | comment | added | James Jenkins | @AaronBertrand RE:"but what actual problem do you have? Do you have one?" that is exactly what I want to be able to find out. When you run a 10 minute mile and your pulse is high, you are not worried. But when you are sitting on the couch and your pulse is high, you also want to check a couple of other things like your BP and Respirations. If/when you check the PLE vital signs what else do you check? | |
| Apr 12, 2018 at 17:25 | answer | added | Brent Ozar | timeline score: 15 | |
| Apr 12, 2018 at 17:23 | comment | added | Aaron Bertrand | Well, you could also say most people think a 4 minute mile is good. If I ever have a 10 minute mile I'll be ecstatic. 350 might be perfectly normal for a given workload, and it might actually be great for a LOT of workloads. If you really want to know about PLE (and Brent explains exactly why you shouldn't focus on that number, especially in isolation), you need to teach SCOM to alert on drastic changes based on your specific workload, not some magic number it pulled from a blog post in 2002. Your PLE is lower now than it was all week, but what actual problem do you have? Do you have one? | |
| Apr 12, 2018 at 17:20 | comment | added | James Jenkins | @AaronBertrand :) "Most can also". But yeah no absolutes here. | |
| Apr 12, 2018 at 17:19 | history | edited | MDCCL | CC BY-SA 3.0 |
Modified title.
|
| Apr 12, 2018 at 17:19 | comment | added | Aaron Bertrand | PLE of 350, why does everyone have to agree that this is not good? For some workloads this is perfectly fine and in fact expected. Having a magic threshold like 300 and alerts keyed off of that is the part that's not good IMHO. | |
| Apr 12, 2018 at 17:16 | history | asked | James Jenkins | CC BY-SA 3.0 |