Yep, my tests were with bluestore. Probably best to just create a
tracker ticket for this and attach the logs for Sam.
Mark
On 5/5/22 08:17, Radoslaw Zarzynski wrote:
Hello!
I guess Mark is testing with BlueStore while you went with SeaStore:
#5 0x00007fe543b6adb5 in abort () from /lib64/libc.so.6
#6 0x0000000000cd6085 in crimson::os::seastore::SeaStore::on_error
(t=...) at /home/ceph/src/crimson/os/seastore/seastore.cc:970
#7 0x0000000001b52707 in
crimson::os::seastore::SeaStore::<lambda(auto:114&,
auto:115&)>::<lambda()>::<lambda(auto:116)>::operator()<std::error_code> (
this=<optimized out>, e=...) at
/home/ceph/src/crimson/os/seastore/seastore.h:226
How about giving a try to / retesting with the former one?
Regards,
Radek
On Thu, May 5, 2022 at 3:03 PM 韩峰哲 <hanfengzhe@xxxxxxxxxx> wrote:
Hello !
When I create 7 crimson-osd, the cluster is normal.
Next, I create 8 crimson.
The pg_1.0 of .mgr pool is created with osds(1, 0, 2). When osd.7
joins, pg_1.0{1,0,2} change to pg_1.0{7,0,2}, and then, osd.7 crashed.
I am confused that Why the pg_1.0 is changed when osd.7 joined the
cluster and why crimson-osd crashed.
I pulled master branch today. The attachments are log and backtrace.
------------------------------------------------------------------
From:Mark Nelson <mnelson@xxxxxxxxxx>
Send Time:2022年4月25日(星期一) 21:56
To:dev <dev@xxxxxxx>
Subject:Re: Test crimson-osd and coredump
Also, fwiw I am testing crimson-osd right now with 60 OSDs and
1024
PGs. There are plenty of other problems, but not this one.
:) Logs,
version, and core dump would all potentially be useful! Thank you.
Mark
On 4/25/22 02:39, Radoslaw Zarzynski wrote:
> Hello!
>
> First of all, thanks for your testing!
> There shouldn't be a hard limit. Could you please provide
backtraces / logs
> from the crashes you're seeing?
>
> Regards,
> Radek
>
> On Mon, Apr 25, 2022 at 8:20 AM 韩峰哲 <hanfengzhe@xxxxxxxxxx>
wrote:
>>
>> I am testing crimson-osd by vstart, the code is master
branch which cloned last week.
>>
>> When I start more than 3 crimsons, if the PG number per
crimson is equal to or more than 8, crimson process will
coredump when running "rados bench ********" everytime, i.e 4
crimsons and 32 pgs in RBS pool.
>>
>> Is there any PG restrict or other restricts when using
crimson-osd ?
>>
>> _______________________________________________
>> Dev mailing list -- dev@xxxxxxx
>> To unsubscribe send an email to dev-leave@xxxxxxx
> _______________________________________________
> Dev mailing list -- dev@xxxxxxx
> To unsubscribe send an email to dev-leave@xxxxxxx
_______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx
_______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx
_______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx
_______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx
