Ive tried to post on this topic before .. this odd “Tx-DROP” counter on my CRS3X switches.. I see them all the time.
Now in this case I do have a 1 gig internet connection that im Accepting on a NetFiber Crs310-in on the copper port, from what I understand it is connected to the switch chip and I can take advantage of hardware offload. And from there I am sending a 1 gig link to my CRS326-24SFP+. Hits my CCR2116. Now connecting from the same CRS326 I have now vlan’d from the CCR2116 I have connected my computer using a Copper module for testing.
Now I know this is a bit confusing but I made a map to maybe help understand how everything connects.
Now when doing speed test from the computer I get INSANE TX-DROP counters. I dont understand why. Even after resetting the counters and doing more conservative iperf3s-.(not hitting the 1gig interface limit, this happens even when doing 400m). the drops still show up.. I mean into the Millions and super fast. the network just got set up so I dont have really real world issue yet. but I mean Im desperate to understand what that is..
Copper SFP is the Mikrotik S+RJ10
Ive written the ROS version as a text over each device.
I went to each and set channel to stable and upgraded to the one it chose.
Any insight would GREATLY be appreciated..
Thought id add the iPerf.. happens in the upload direction.
iperf3 -c myiperf.server -u -b 200m
and tried with multiple streams -P 2
Yet these are my results? I dont get it..
SO the hell are these drops?!
I know I keep writing on my own post.. but does this maybe have something to do with the 10 gig links and the 1gig wan. I think ive read somewhere that, those things could present issues sometimes. I dont I feel braindead
Keep your hair on!
I reported this 3 years and 4 months on ago for a set of CRS328 switches and got this back.
“Regarding the CRS328, we have discovered in our labs that the “tx-drop” increases the counter although no packets are being dropped, on the other hand, it also does not increase the counter when the packets are actually being dropped (e.g. when congestion is actually happening). To summarize, the counter is not showing correct data and cannot be relied on. We will try to fix this in upcoming RouterOS releases.”
I checked in a year later and still not fixed then. I haven’t looked since. Methinks it’s still a thing.
I think its like RX Drops - when a vlan tag is passed thru the device that it doesnt know about. Turn on torch or packet sniffer and run your same tests and see if it stops.
I have just looked at the particular switch in question. It was recently upgraded from ROS6 which it ran from the start 3 years ago to ROS7.13.2 and the port I was looking at which sends all security camera streams to the main recorder has 0 TX-DROPs.
So maybe they fixed it. This port only has untagged traffic and a PVID set. Check your ROS version - is it the latest ROS7?. Make sure the firmware version matches too. Post your config here before you send supout.rif’s to Mikrotik support.
On some of the CRS326’s I have I dont know if it’s the revision, but the latest stable is 6.49. and I might be able to upgrade. but not being on stable scares me. I did try and switch the link negotiation and force it to 1gig to my router.. which matches the negotiation where I 1gig connected to the internet connection. Seems to of helped. im going to check again here tomorrow with more intense stress test
There is no harm in using ROS7. ROS6 is a legacy version now and any new fixes will only appear in 7.
Export your config (don’t forget the show-sensitive flag), save it off the switch and upgrade. See what happens. You can downgrade the switch by wiping it, installing your old ROS6.49.10 image and reimporting the config.