hello, everyone, I am running influxdb v2.3. i found my influxdb is so slow because my colloegue begin to integrate 6 device to push data in. And i check my system by use top and iostat. i found there are no a lot of wirte but very loads of read. it consume the disk 100% all the time. I don’t know what the influxdb is doing backward.
Is anyone could help me ? I am appreciate very much.
(my disk is the IBM storage but not ssd,but we just only connect 3 device.)
I am puzzled by two things you say:
-
“my colleague began to integrate 6 device to push data in” (so, that’s
writing to the DB) -
“I found there are not a lot of writes but loads of reads” (so, that’s
reading from the DB)
So, it is not at all clear to me whether your problem is too many writes (from
your colleague) or too many reads (from wherever).
I suggest you give us some clear numbers to work with:
a) what total quantity of data do you have in the DB currently?
b) how much data is being added per second / minute / hour / whatever?
c) are you writing lots of little bits of data, or fewer large lumps of data?
d) what is the data retention policy (ie: is old data being deleted from the
DB)?
e) how complex is the data - give us an example (obfuscated if anything is
confidential) of what you are storing
Finally, what are youi doing with this data? What is reading data out of
Influx DB (and, how much / how often)?
Antony.
hello pooh thanks for your help.
a) i check my disk it’s 85G data in the DB now.
b) i ask my colleage, he said about 600 entrys every 30 second.
c) a little bits of data
d) i have two bucket below, the devices wirte the data to “heat_treatment_params”, and this bucket 's retention is 30days. then i create a “heat_treatment_params_replica” to replica from “heat_treatment_params”. this bucket’s retention is infinite to save data forever. why I create tow bucket like that because there is no increment backup for influxdb2.0. So i could backup the “heat_treatment_params” every one day beacuse the data is just only 30days. and I backup the “heat_treatment_params_replica” once a month for data security.
876063dc27033e1e heat_treatment_params 720h0m0s 24h0m0s af34aa784426d070 implicit
c5e426e179ea1ad9 heat_treatment_params_replica infinite 168h0m0s af34aa784426d070 implicit
e) the example data: the msgid is nonsense, so i ask my colleage to delete next time.i guess this msgid will go against the classification.
_start 2022-12-28 08:19:29 GMT +8
_stop 2022-12-28 11:19:29 GMT+8
_time 2022-12-28 10:20:00 GMT +8
_value 25.76
_field timeDiff
_measurement 27036764-a111-11ec-8aeb-240617a558d5
equipmentId 27036764-a111-11ec-8aeb-240617a558d5
msgId 76584
tagId 283b71b2-a111-11ec-984e-24
d) we use the grafama to show some graph from influxdb.
but we try to stop the grafama ,it seem not the grafama’query cause the problem.
f) yesterday I reboot the server at 12:00. when the service start it begin to occupy all the disk IO.and till 22:00, it return to normal.so I dont know if there is something task like data compression is working backward?
