Linux Operating System (MOSC)

MOSC Banner

Large IO contention on powerful storage

edited May 5, 2022 3:02AM in Linux Operating System (MOSC) 3 commentsAnswered

On 27 March system (Oracle Linux) has been upgraded to 5.4.17-2136.305.5.3. Also Oracle Database EE to 12.2.0.1.220118. Storage is Pure Storage X20 R3 (40TB)

After over a week of reasonably good performance, starting from 06.04 (Wednesday) large performance decrease has been noticed (but workloads stayed the same/similar). In everyday critical period 07:00 – 09:30 we observe many wait events (latch: cache buffers chains and free buffer waits mainly) and, above all, IO delays. OK, huge workload.

But at the same time strange system behaviour: while we observe OEM “Latency For Synchronous Single Block Reads” growing up (to 10ms from the usual 0,5ms) in the same time we see plenty (thousands!) of kworker system process. Result of Tanel Poder’s Process Snapper :

Howdy, Stranger!

Log In

To view full details, sign in to My Oracle Support Community.

Register

Don't have a My Oracle Support Community account? Click here to get started.

Category Leaderboard

Top contributors this month

Instructions for posting

×

1. Select a discussion category from the picklist.


2. Enter a title that clearly identifies the subject of your question.


3. In the body, insert detailed information, including Oracle product and version.


Please abide by the Oracle Community guidelines and refrain from posting any customer or personally identifiable information (PI/CI).


Cloud / Fusion customers - Our Cloud community has moved! Please go to Cloud Customer Connect .


New to My Oracle Support Community? Visit our Welcome Center

MOSC Help Center