Re: [BUG] active zones exceeded error with max_open_zones

On 4/24/25 10:01, Sean Anderson wrote:
> On 4/24/25 02:13, Shinichiro Kawasaki wrote:
>> On Apr 23, 2025 / 13:11, Sean Anderson wrote:
>>> Hi,
>>>
>>> I'm getting an "active zones exceeded" error when running fio with
>>> --rw=randwrite mode:
>>>
>>> # fio --bs=4k --rw=randwrite --norandommap --fsync=1 --number_ios=16384 --name=flushes --direct=1 --zonemode=zbd --max_open_zones=1978 --filename=/dev/my_zone_dev
>>> flushes: (g=0): rw=randwrite, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=psync, iodepth=1
>>
>> Hi,
>>
>> The block size is 4k. And according to the blkzone report, it looks like the gap
>> between the zone size and zone capacity is 512b. So, I guess the fio can not
>> fill the gap by 4k writes. It looks likely this unalignment between the gap and
>> the block size left unwritten small remainders in many zones, then many zones
>> are kept open and the device exceeded the max active zone limit. Based on this
>> guess, I suggest to try with 512 byte block size to align it with the gap.
> 
> Same issue. And fio seems to be capable of working with the gap in --rw=write mode.
> 
>>> fio-3.39
>>
>> Recently, I contributed a fix which handles the case many small remainders are
>> left by random write workload to zoned block devices [1]. The fix was upstreamed
>> after the fio version 3.39. I also suggest to try out the latest fio code with
>> 4k blocksize.
>>
>> [1] https://github.com/axboe/fio/commit/e2e29bf6f8300186d267fa46a7b266d14d174575
> 
> Yeah, I saw that, but it didn't appear to be related to this issue. I will try it
> out tonight.

Same issue. I will look into this more this weekend.

--Sean