RE: trim entries leaked?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Remember trims are optionally performed by the drive and the drive at any given time can accept the trim and throw the command away and the original data stays.

If the original data is trimmed, then data pattern requirements come into play (i.e. zeros, etc.)

Jeff
 
Jeff Wolford                       Email: Jeff.Wolford@xxxxxxx
Distinguished Technologist
HPE Servers
    Teams: (832) 502-5749
Hewlett Packard Enterprise

-----Original Message-----
From: Matthew Eaton <m.eaton82@xxxxxxxxx> 
Sent: Friday, October 6, 2023 2:53 PM
To: fio@xxxxxxxxxxxxxxx
Subject: trim entries leaked?

Hi fio group,

I'm trying to get a workload doing concurrent writes, verifies, and/or trims. I've narrowed down my job file to this:

[verify]
ioengine=psync
direct=1
size=5g
bs=4k
rw=randwrite
numjobs=1
iodepth=1
verify=crc32c
verify_backlog=1
trim_percentage=50
trim_backlog=1
filename=/dev/nvme0n1

With this job writes are read/verified but there are no trims. If I set backlogs to 0 I get this message before the verification phase:

fio: 654981 trim entries leaked?

I'm using the latest fio as of today, fio-3.35-134-g6f9c.




[Index of Archives]     [Linux Kernel]     [Linux SCSI]     [Linux IDE]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux SCSI]

  Powered by Linux