Sometimes I temporarily install Pakon, but the problem of night 2-hour writing to the microSD in the / srv / storage function still persists. I wonder if my problem is unique or not and if it is a feature or a bug of Pakon. Gitlab did not react constructively.
TOS5.3.2 Turris Omnia
https://forum.test.turris.cz/t/night-load-of-the-router-pakon/15563
https://gitlab.nic.cz/turris/pakon/-/issues/22
I was looking on the issue on gitlab and assigned to one of the colleagues to verify it. In general, it is expected. Pakon aggregates data at 2 o’clock. The question is what level is bug and what level is expected. In short if you have a lot of network traffic then it is expected that archiving takes a lot of time. It takes even longer time if storage medium is slow.
You should in general use Pakon with some faster storage such as USB 3.0 drive. SD card is slower and it is also expected that it is going to be wear it out pretty quickly this way. This gets more important if you have a lot of traffic (this is less about volume and more about variety of it.
The card https://www.alza.cz/kingston-microsdhc-8gb-industrial-sd-adapter-d6713538.htm?o=2 front USB 3 port with adapter Kinston micro reader FCR-MGR2 (USB2) …
… is used as Storage /srv/ ( USB 2 write <35 MB/s ??). The folder /srv/pakon has a size of 11012 k (ekv. 11 MB).
The amount of data does not indicate the need to copy data for 2 hours. The indicated write speed of the used storage (mentioned adapter and microSD) in the PC deskotop is 25 MB. If I could get a reasonable USB3> microSD reader (https://www.alza.cz/kingston-mobilelite-plus-uhs-ii-microsd-reader-d5787696.htm?o=4) I can try this variant, but I don’t give much hope. Pakon’s data just isn’t that big
Is there a tool to record this special load event of CPU and write (Pakon debug )?
Aggregating data is not the same as copying. It takes more than just copying it. This runs through the database and removes and adds data from it and to it. The slow storage is going to humper that.
I think that you would get way more from a small USB 3.0 flash drive than this slow reader with SD card.
I have a bad experience many years with flash drives. For It I chose industrial SD. I will consider your advice.
EDIT:
So I put the SanDisk Ultra Flair 32GB into the test. We’ll see what the results will be, the first night looks promising
Thank you @cynerd for heading in the right direction, tonight’s night reading on the 1 hour frame proved that the previously problematic Pakon operation will take place in about 10 minutes. It can also be seen from the graphs that the CPU and Load are at the same level, but the Write to /sdb/ has increased from 250kB/s …(15 operations/s) to 5 MB/s …(150 operations/s)
This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.
