Ilya Mashchenko

Hi, @k0ste. > ip link set X down just changes the interface operstate (/sys/class/net/X/operstate). There is no problem if /sys/class/net/X/operstate is readable. --- How it works now: - Netdata reads...

Hi, @k0ste. We plan to implement the following fix: if the port is physical we assume it always should have speed, duplex, carrier, and operstate. So for physical ports, we...

Hi, @k0ste. Can you test the [fix](https://github.com/netdata/netdata/pull/13850)?

Hey, @romain-dartigues. Are you going to finish your PR?

@shyamvalsan I don't think we are at the point where we can add anomaly alerts by default: - not tested. - no certain action from a user perspective (anomalous and...

I don't see our latency implementation (PING commands) is _alertable_: - I have no clue about the baseline. - I guess you had a local instance in mind suggesting 200ms...

@shyamvalsan I just started a few Redis containers (using the official image, redis_version:7.0.4) and mem_fragmentation_ratio is 12-13 (no load at all).

It is used_rss / used_memory ```bash docker run --name some-redis -d redis docker exec some-redis redis-cli INFO MEMORY | grep fragmentation ```

@shyamvalsan ok, the description > Note that this doesn't only includes fragmentation, but also other process overheads (see the allocator_* metrics), and also overheads like code, shared libraries, stack, etc....

Added 2 and 4. - mem_fragmentation: not an accurate metric according to my quick research. At the same time can't say that _allocator_frag_ratio_ is the way to go. - **hit**rate:...