Hi! Any chance to get it for Christmas? 
I am also wondering on this. I can see the changelog shows v2.9 when checking for auto updates, but it doesn’t seem to be available. RouterOS upgraded SwOS to 2.8p with build date 7/14/18 (yes, there’s a p), but this ‘p’ version doesn’t show in the change log or the site. Trying both auto and manual updates fail to install the 2.8 (no ‘p’) with build date of 10/4/18. They act like they are updating, but come back up with 2.8p again.
Please advise as I’ll have to RMA these switches if I can’t find a solution soon.
Thank!
Using the SFP instead of an ethernet port for the uplink, solved my overrun (overflow on RouterOS) problem.
Same problem →
CSS326-24G-2S+RM, SwOS 2.9 with 10G SFP+ upLink to
CRS317-1G-16S+, SwOS 2.7
with disabled flow control :
on CSS326 lots of Rx Overruns
on CRS317 lots of Rx Pauses
with enabled flow control :
lots of Tx/Rx Pauses on both switches
no Tx/Rx Overruns on both switches
Traffic about 1,5 Gbps between two switches - most of it with MTU 9000.
Any ideas - i seem to be lost on this…
I’ve given up on Mikrotik switches. I have two CSS326 connect through sfp+ fiber (S+85DLC03D in both ends). A lot of RX overruns, some RX FCS Errors and also on many of the other ports, not only the sfp ports. I’ve now moved almost all equipment to HP switches and only using a two port LAG to the HP switches and the fiberlink. Also enabled flowcontrol on SFP port. Now at least I have no errors but a lot of TX and RX pauses. With the equipment I have I can get ~3,5 Gbps throughput (can’t put more load on the link with current setup). The HP switches reports no errors. Running swos 2.9 and have had the same results with earlier versions of swos. My conslusion is that either swos is broken and/or the hardware (CSS326) doesn’t have enough capacity (buffers or whatever). So those running the CRS326 with Routeros seems to get it working why I am leaning towards swos problem. I hope the CRS305 and CRS309 are working better since I’ve been thinking of getting some of those but right now I’m looking at other brands.
Same here!
Lot of tx/rx overruns on 10G port (SFP+ to CRS317) or pauses when flow control is enabled.
I’ll wait till next SWoS release and if it is not fixed will throw away and will declare it as another cheap junk.
You are welcome to send them to me. I have been very happy with my CRS326 (in SWOS) and two CSS326 switches and would be happy to have a spare I have had no problems with the three, but I’m not trying to connect at 10Gb/s - max of 1Gb/s.
if you thing this is normal - errors tab :
Rx Overruns Tx Pauses
SFP tab:
Not my experience. Just looked at one of mine which shows an uptime of 199 days. There is 1 RX Overrun and 0 TX pauses. As I said, nothing faster than 1Gb/s…
But with 199 days uptime You are running 2.8, right? I upgraded from 2.8 to 2.9 (CSS326), and saw the increase in Rx Overrruns and Tx Pauses. They are too frequent to my liking, but I don’t think they have a real impact on network speed. At least, not for me.
Truth be told, although I use a 10GB trunk, the traffic is light - so maybe it is worse with heavier loads.
=== EDIT ===
10Gb trunk. That typo was killing me!
Better now. 
=== /EDIT ==
Good catch. That one had not been updated. I had forgotten that I had not updated that one yet. I have a bunch of network changes coming up in a few days because of a change in one of my internet connections, so that will be a good time to update that one. A couple routers will also be updated at that time too…
i think its a HOL (Head of LIne) Blocking issue
in mi case it is enough that an download interface (toward clients) negotiate at 100m or 10m is saturated to start to increase dramatically overrun counters on SWos, if running Routeros the count is called overflow in the interface receiving all the traffic from main router
tested with:
CSS 326
CRS 326
CRS 328
That means - we have that problem when traffic passes between interfaces with different speed. Most of people, I think, byes those models to extend 10G networks to old 1G equipment and it’s not work…
I receive overruns on 10G eth (up-link) only.
agree with you, that’s the issue, when a down-link port gets saturated the up-link port suffer from overrun/overflow
in my case im only using 1g ports and the issue is present, that imply is not exclusive of 10g uplink vs 1g downlink scenario
i was recommending this switches but from now i am stopping mi customers (mostly isp’s) from buying until this issue gets resolved
Same issue on my CSS. Overruns on 10g trunk port and pauses on CRS side running routeros.
Cannot say if it affects the experience a lot, haven’t catched any packet loss and the forwarding is not too bad, as I was able to record files to a remote PC at 95Mb/s through an SMB protocol. Haven’t tested the actual SFP+ speeds yet as it’s only between the network devices. The end users are maximum 1gbit.
Would try to do more speed testing to see if I’m able to push at least 1g simultaneously between multiple PC’s.
Отправлено с моего Mi A1 через Tapatalk
Same here… Only overruns, only at one interface 10G
Interface traffic most time is passing 2Gbps+
Linked this CSS326 with CRS317… with s+da0001
When Flow Control is enabled, we see RX Pauses.
When Flow Control is disabled, we see Rx Overruns
Seems the problem is with CSS326, because, we moved one 10G interface from CSS326 to the CRS317… and it is running without erros.
No change on modules was made.
Really strange behavior.
I have this same issue 2.10 I am wondering if anyone can say if they have this issue using Mikrotik SFP+ transceiver and any brand of fiber. I am starting to wonder if its a DAC issue. I can see this between my CSS3266-24G-2S+RM and my CRS309-1G-8S+ using a 10m Cisco active Twinax, as well as between my CRS309-1G-8S+ and my ESX host where there is like for like bandwidth and like for like a machine. When pushing 5gbps+ between ESX host that each have 2 10G i interfaces in an active LACP bond between them and the CRS309-1G-8S+, with 3m cisco Twinax.
Even pushing a solid 10gbps and getting this pause incrementing like my utility meter its not a huge impact on latency or bandwidth, but it is a problem if you had more sensitive setup than my home lab.
very rare
with 2.10 solved the issue on most scenarios
This is why I am wondering if its the DAC’s. On the interface labeled Core94 I can get seeing this type of error as its a 10g link that downstream has to exit 1g link either to a lan device or the fiber internet. However on the esx it doesn’t make much sense they each have bonded 10g pairs and I can see in the stats its load balancing pretty well for tx no so well for rx, to generate these errors here I ran a iperf for 3 minutes from esx03 to esx01 and a second iperf from esx03 to esx02. Its very interesting the esx03 got no errors while the other two host got errors, again leading me to ask of three identical network cards, going to the same switch what is different and the only thing is the dac’s.
I honestly can say for my current needs its more of a visual issue seeing the errors than an actual issue I notice. But at the same time if swapping out dac’s for optical would actually solve the problem then thats what I might have to do.
edit: On closer inspection I would expect it to be TX pause not RX pause on the core94 since its was generated running iperf to my colo so sending from a device behind this switch to an internet location limited to 1g
