hey,

 

I'm having some trouble with my ssh connection to an armbian based server. It feels like a heavy lag within the filesystem, sometimes it takes up to 20 seconds to open a file with vim or use cd with tab-completion. But it is not always there, it feels more like some heavy write operation occurs spontaneously. I start digging but I have no clue why this happens. But at least I found the most disk I/O processes which are influxdb and mysql.

 

[16:54|root@odroidhc1 ~]# cat /syscat /sys/fs/ext4/mmcblk0p1/session_write_kbytes 
118811040

[16:55|root@odroidhc1 ~]# uptime
 16:56:03 up 10 days,  3:53,  3 users,  load average: 0,87, 0,64, 0,59
 
[17:15|root@odroidhc1 zram-config]# iostat
Linux 4.14.150-odroidxu4 (odroidhc1)    12.01.2020      _armv7l_        (8 CPU)

avg-cpu:  %user   %nice %system %iowait  %steal   %idle
           2,03    0,44    0,93    0,85    0,00   95,75

Device             tps    kB_read/s    kB_wrtn/s    kB_read    kB_wrtn
loop0            24,86        12,44         0,00   10932869        668
mmcblk0           8,55        90,30       135,32   79391237  118968040
zram0             0,88         1,54         1,97    1349836    1735284
zram1             2,35         3,83         5,58    3369880    4903220
sda               1,10       110,95        47,61   97549568   41860980
dm-0             24,86        12,43         0,00   10931824        668


[16:36|root@odroidhc1 ~]# cat /etc/fstab 
# <device>  <dir>   <type>  <options>   <dump>  <fsck>
UUID=8a464498-f41a-4de4-9a05-52e636c8135e / ext4 defaults,noatime,nodiratime,commit=600,errors=remount-ro 0 1
UUID=92d36a67-0322-42cf-9a4c-81a0563f58c9 /mnt btrfs defaults,nofail 0 1

tmpfs /tmp tmpfs defaults,nosuid 0 0

tmpfs   /var/www/tmp    tmpfs    defaults,nofail,noatime,nosuid,uid=33,gid=33,mode=0755,size=100m    0 0
tmpfs   /var/log/audit    tmpfs    defaults,nofail,noatime,nosuid,mode=0755,size=100m    0 0

[16:37|root@odroidhc1 ~]# swapon
NAME       TYPE        SIZE USED PRIO
/dev/zram1 partition 998,5M 246M    5

However, the value (session_write_kbytes) is constantly rising and not within the commit time delay from fstab file. The mariadb is set to fsync and not o_direct, it should respect the commit entry?

However (118811040 KBytes == 116026 MBytes == 113 GBytes) within 10 days? It seems a lot for a single user nextcloud instance? The influxdb takes like 6 values every minute from 5 inputs.

 

Did someone optimize the mariadb setup for a sdcard?

 

I had a problem with ext4 and data being committed every 30 seconds and changing commit interval in fstab made no difference to fix it I modified:

 

/proc/sys/vm/dirty_expire_centisecs

 

to 500 (5 seconds). 30 seconds is default

 

You can monitor it with:

 

 cat /proc/meminfo | grep "Dirty" -A1

 

thank you, I give it a try.

Currently I switched to the HDD, the amount of written data is enormous, like about 200GB for mysql and 150GB in two hours! Luckily the responsiveness raised a lot.