Cloudflare local CDN unavailability

cavedog

Honorary Master
Joined
Oct 19, 2007
Messages
26,965
Reaction score
13,126
Location
PTA
@rpm

So on 11th of September Cloudflare switched off if I can say that the local CDN for all cloudflare services not on their enterprise package.

I tried to get info regarding this but it's not forth coming. Currently all traffic is routed by the ISP to the next CDN with some going to London and others like Afrihost routing to Kigali Rwanda....

Any solution or info from anyone?

@websquadza @waylander
 
@rpm

So on 11th of September Cloudflare switched off if I can say that the local CDN for all cloudflare services not on their enterprise package.

I tried to get info regarding this but it's not forth coming. Currently all traffic is routed by the ISP to the next CDN with some going to London and others like Afrihost routing to Kigali Rwanda....

Any solution or info from anyone?

@websquadza @waylander

Hi,

We are suffering from the same issue. Logged a support ticket with them. This is the lovely response i get

"I have checked with my operation team and I'm afraid that there will not be any immediate progress in at least the next few weeks as they are working on upgrading/expanding the data centers in ZA.
The only option will be to upgrade to the Enterprise plan."

Enterprise starts at 3000$ according to there sales team. Which are in China & have never heard of Cloudflare ever having such a issue before.

Takealot & MyBroadband sites suffering from the same issue.
 
@rpm

So on 11th of September Cloudflare switched off if I can say that the local CDN for all cloudflare services not on their enterprise package.

I tried to get info regarding this but it's not forth coming. Currently all traffic is routed by the ISP to the next CDN with some going to London and others like Afrihost routing to Kigali Rwanda....

Any solution or info from anyone?

@websquadza @waylander

Some feedback from Cloudflare here: they experienced an issue with the transit provider feeding their SA DCs. They’ve been forced to drop all content except for enterprise clients until a solution is found. No ETA.
 
Is that why myBB is laggy?

Yes right now if you host a website in SA or in EU West it is better to switch off CDN if the majority of your traffic is from SA as the routing is not optimal unless the ISP intervenes to fix some routing.

That is the reason I tagged @rpm because an article will be nice and would probably be better if they deselect the CDN option in cloudflare until they get DDoS or Cloudflare get the transit fixed.
 
Some feedback from Cloudflare here: they experienced an issue with the transit provider feeding their SA DCs. They’ve been forced to drop all content except for enterprise clients until a solution is found. No ETA.
@websquadza - can you ask Carl to reply to my email?
 
@AfriMan @Afrihost-Gian Please can you talk to NOC to force cloudflare routing to London instead of Kigali. Would be so much better.

They could probably reroute outbound traffic to the EU. But unless Liquid stops advertising client prefixes to Cloudflare at its next closest node, return traffic will still be from Kigali. Not to mention the massive increase in latency with packets going up one coast and down the other.
 
They could probably reroute outbound traffic to the EU. But unless Liquid stops advertising client prefixes to Cloudflare at its next closest node, return traffic will still be from Kigali. Not to mention the massive increase in latency with packets going up one coast and down the other.

Yeah I will query with them.

The issue with Kigali is not latency. Going to Mtunzini over seacom then lands in Nairobi then goes inland to Kigali. Why not just use Nairobi CDN rather. I'm doubting the capacity of Kigali because latency is still nice and low. Less than 90ms.

I'm surprised Cloudflare is leaving it because the traffic must have spiked like crazy for Kigali. Afrihost, Axxess and Webafrica all using Liquid which is going there.
 
Is that why myBB is laggy?
@rpm

I'd say "laggy" is a massive understatement. With over 50% packet loss on Liquid Telekom it is near unusable.

Feel free to cry ...

~$ tcptraceroute mybroadband.co.za
Selected device eth0, address ####, port 38972 for outgoing packets
Tracing the path to mybroadband.co.za (104.20.10.169) on TCP port 80 (http), 30 hops max
1 10.0.0.1 0.484 ms 0.289 ms 0.254 ms
2 * * *
3 jhb-up3.ip.adsl.co.za (169.1.5.29) 5.212 ms 3.782 ms *
4 jhb-in1.ip.adsl.co.za (169.1.5.42) 4.920 ms 4.137 ms 4.207 ms
5 jhb-net1.ip.adsl.co.za (169.1.5.52) 6.939 ms 5.538 ms 4.441 ms
6 ae-21.lza.pr1-jhb.liquidtelecom.net (77.246.59.254) 5.226 ms 4.327 ms 4.283 ms
7 46.17.232.152 74.858 ms 74.788 ms 74.369 ms
8 te-0-2-0-0.lke-p2-msa.liquidtelecom.net (77.246.56.246) 79.802 ms 84.259 ms 123.633 ms
9 et-0-1-11.lke-p1-msa.liquidtelecom.net (197.155.94.15) 75.309 ms 76.907 ms 75.766 ms
10 hu-0-3-0-3.lke-p1-nbi.liquidtelecom.net (197.155.94.216) 79.218 ms 79.167 ms 78.752 ms
11 te-0-0-1-1.lug-p2-kpl.liquidtelecom.net (197.155.94.69) 79.345 ms 79.402 ms 78.866 ms
12 te-0-0-1-2.lrw-pe1-kgl.liquidtelecom.net (197.155.94.163) 78.774 ms 80.499 ms 79.141 ms
13 41.223.224.175 76.090 ms 80.142 ms *
14 104.20.10.169 [open] 79.105 ms 75.024 ms 73.670 ms


All good so far ...
~$ ping 197.155.94.163
PING 197.155.94.163 (197.155.94.163) 56(84) bytes of data.
64 bytes from 197.155.94.163: icmp_seq=1 ttl=245 time=79.7 ms
64 bytes from 197.155.94.163: icmp_seq=2 ttl=245 time=78.4 ms
64 bytes from 197.155.94.163: icmp_seq=3 ttl=245 time=79.6 ms
64 bytes from 197.155.94.163: icmp_seq=4 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=5 ttl=245 time=78.1 ms
64 bytes from 197.155.94.163: icmp_seq=6 ttl=245 time=78.4 ms
64 bytes from 197.155.94.163: icmp_seq=7 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=8 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=9 ttl=245 time=78.2 ms
64 bytes from 197.155.94.163: icmp_seq=10 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=11 ttl=245 time=78.1 ms
64 bytes from 197.155.94.163: icmp_seq=12 ttl=245 time=78.1 ms
64 bytes from 197.155.94.163: icmp_seq=13 ttl=245 time=78.3 ms
^C
--- 197.155.94.163 ping statistics ---
13 packets transmitted, 13 received, 0% packet loss, time 12017ms
rtt min/avg/max/mdev = 78.100/78.513/79.743/0.529 ms


~$ ping 41.223.224.175
PING 41.223.224.175 (41.223.224.175) 56(84) bytes of data.
64 bytes from 41.223.224.175: icmp_seq=3 ttl=53 time=75.0 ms
64 bytes from 41.223.224.175: icmp_seq=5 ttl=53 time=74.5 ms
64 bytes from 41.223.224.175: icmp_seq=6 ttl=53 time=74.3 ms
64 bytes from 41.223.224.175: icmp_seq=7 ttl=53 time=92.6 ms
64 bytes from 41.223.224.175: icmp_seq=10 ttl=53 time=74.3 ms
64 bytes from 41.223.224.175: icmp_seq=12 ttl=53 time=74.2 ms
64 bytes from 41.223.224.175: icmp_seq=16 ttl=53 time=95.9 ms
64 bytes from 41.223.224.175: icmp_seq=17 ttl=53 time=74.5 ms
64 bytes from 41.223.224.175: icmp_seq=19 ttl=53 time=74.3 ms
64 bytes from 41.223.224.175: icmp_seq=20 ttl=53 time=74.2 ms
^C
--- 41.223.224.175 ping statistics ---
21 packets transmitted, 10 received, 52% packet loss, time 20027ms
rtt min/avg/max/mdev = 74.235/78.429/95.910/7.957 ms

41.223.224.175 lookup

% No abuse contact registered for 41.223.224.0 - 41.223.225.255

inetnum: 41.223.224.0 - 41.223.225.255
netname: XIOCOM
descr: Xiocom infrastructure and client IP addreses
country: RW
admin-c: AA110-AFRINIC
tech-c: AA110-AFRINIC
status: ASSIGNED PA
notify: [email protected]
mnt-by: LIQUID-TOL-MNT
changed: [email protected] 20080707
changed: [email protected] 20140422
source: AFRINIC
parent: 41.223.224.0 - 41.223.227.255

person: Andrew Alston
address: Block A, Sameer Business Park,
address: Mombasa Road,
address: Nairobi
address: Kenya
phone: tel:+254-20-5000000
e-mail: [email protected]
nic-hdl: AA110-AFRINIC
mnt-by: AA110-MNTR
changed: [email protected] 20170208
source: AFRINIC

% Information related to '41.223.224.0/24AS30844'

route: 41.223.224.0/24
descr: Maintainer Liquid Telecommunications Operations Limited
origin: AS30844
org: ORG-LTOL1-AFRINIC
mnt-lower: LIQUID-TOL-MNT
mnt-by: AFRINIC-HM-MNT
changed: [email protected] 20180907
changed: [email protected] 20180910
source: AFRINIC

organisation: ORG-LTOL1-AFRINIC
org-name: Liquid Telecommunications Operations Limited
org-type: LIR
country: MU
address: 10th Floor,
address: Raffles Tower,
address: 19 Cybercity
address: Ebene
phone: tel:+230-466-7620
phone: tel:+254-733-222204
phone: tel:+263-8677-033306
phone: tel:+254-731-033754
e-mail: [email protected]
e-mail: [email protected]
e-mail: [email protected]
e-mail: [email protected]
e-mail: [email protected]
admin-c: CM53-AFRINIC
admin-c: AS116-AFRINIC
admin-c: RD10-AFRINIC
admin-c: DH19-AFRINIC
tech-c: CM53-AFRINIC
tech-c: AS116-AFRINIC
tech-c: AA110-AFRINIC
tech-c: DH19-AFRINIC
tech-c: DV5-AFRINIC
tech-c: KR12-AFRINIC
mnt-ref: AFRINIC-HM-MNT
mnt-ref: LIQUID-TOL-MNT
notify: [email protected]
mnt-by: AFRINIC-HM-MNT
changed: [email protected] 20140219
changed: [email protected] 20140305
changed: [email protected] 20140327
changed: [email protected] 20140328
changed: [email protected] 20150303
changed: [email protected] 20160722
changed: [email protected] 20171215
changed: [email protected] 20171218
changed: [email protected] 20180214
changed: [email protected] 20180320
changed: [email protected] 20180831
changed: [email protected] 20190617
source: AFRINIC

:X3:
 
They could probably reroute outbound traffic to the EU. But unless Liquid stops advertising client prefixes to Cloudflare at its next closest node, return traffic will still be from Kigali. Not to mention the massive increase in latency with packets going up one coast and down the other.
Also over 50% packets going AWOL - see previous post. :(
 
@rpm

I'd say "laggy" is a massive understatement. With over 50% packet loss on Liquid Telekom it is near unusable.

Feel free to cry ...

~$ tcptraceroute mybroadband.co.za
Selected device eth0, address ####, port 38972 for outgoing packets
Tracing the path to mybroadband.co.za (104.20.10.169) on TCP port 80 (http), 30 hops max
1 10.0.0.1 0.484 ms 0.289 ms 0.254 ms
2 * * *
3 jhb-up3.ip.adsl.co.za (169.1.5.29) 5.212 ms 3.782 ms *
4 jhb-in1.ip.adsl.co.za (169.1.5.42) 4.920 ms 4.137 ms 4.207 ms
5 jhb-net1.ip.adsl.co.za (169.1.5.52) 6.939 ms 5.538 ms 4.441 ms
6 ae-21.lza.pr1-jhb.liquidtelecom.net (77.246.59.254) 5.226 ms 4.327 ms 4.283 ms
7 46.17.232.152 74.858 ms 74.788 ms 74.369 ms
8 te-0-2-0-0.lke-p2-msa.liquidtelecom.net (77.246.56.246) 79.802 ms 84.259 ms 123.633 ms
9 et-0-1-11.lke-p1-msa.liquidtelecom.net (197.155.94.15) 75.309 ms 76.907 ms 75.766 ms
10 hu-0-3-0-3.lke-p1-nbi.liquidtelecom.net (197.155.94.216) 79.218 ms 79.167 ms 78.752 ms
11 te-0-0-1-1.lug-p2-kpl.liquidtelecom.net (197.155.94.69) 79.345 ms 79.402 ms 78.866 ms
12 te-0-0-1-2.lrw-pe1-kgl.liquidtelecom.net (197.155.94.163) 78.774 ms 80.499 ms 79.141 ms
13 41.223.224.175 76.090 ms 80.142 ms *
14 104.20.10.169 [open] 79.105 ms 75.024 ms 73.670 ms


All good so far ...
~$ ping 197.155.94.163
PING 197.155.94.163 (197.155.94.163) 56(84) bytes of data.
64 bytes from 197.155.94.163: icmp_seq=1 ttl=245 time=79.7 ms
64 bytes from 197.155.94.163: icmp_seq=2 ttl=245 time=78.4 ms
64 bytes from 197.155.94.163: icmp_seq=3 ttl=245 time=79.6 ms
64 bytes from 197.155.94.163: icmp_seq=4 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=5 ttl=245 time=78.1 ms
64 bytes from 197.155.94.163: icmp_seq=6 ttl=245 time=78.4 ms
64 bytes from 197.155.94.163: icmp_seq=7 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=8 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=9 ttl=245 time=78.2 ms
64 bytes from 197.155.94.163: icmp_seq=10 ttl=245 time=78.3 ms
64 bytes from 197.155.94.163: icmp_seq=11 ttl=245 time=78.1 ms
64 bytes from 197.155.94.163: icmp_seq=12 ttl=245 time=78.1 ms
64 bytes from 197.155.94.163: icmp_seq=13 ttl=245 time=78.3 ms
^C
--- 197.155.94.163 ping statistics ---
13 packets transmitted, 13 received, 0% packet loss, time 12017ms
rtt min/avg/max/mdev = 78.100/78.513/79.743/0.529 ms


~$ ping 41.223.224.175
PING 41.223.224.175 (41.223.224.175) 56(84) bytes of data.
64 bytes from 41.223.224.175: icmp_seq=3 ttl=53 time=75.0 ms
64 bytes from 41.223.224.175: icmp_seq=5 ttl=53 time=74.5 ms
64 bytes from 41.223.224.175: icmp_seq=6 ttl=53 time=74.3 ms
64 bytes from 41.223.224.175: icmp_seq=7 ttl=53 time=92.6 ms
64 bytes from 41.223.224.175: icmp_seq=10 ttl=53 time=74.3 ms
64 bytes from 41.223.224.175: icmp_seq=12 ttl=53 time=74.2 ms
64 bytes from 41.223.224.175: icmp_seq=16 ttl=53 time=95.9 ms
64 bytes from 41.223.224.175: icmp_seq=17 ttl=53 time=74.5 ms
64 bytes from 41.223.224.175: icmp_seq=19 ttl=53 time=74.3 ms
64 bytes from 41.223.224.175: icmp_seq=20 ttl=53 time=74.2 ms
^C
--- 41.223.224.175 ping statistics ---
21 packets transmitted, 10 received, 52% packet loss, time 20027ms
rtt min/avg/max/mdev = 74.235/78.429/95.910/7.957 ms



:X3:
For me it's completely usable but laggy. Not as bad as you describe it.
 
For me it's completely usable but laggy. Not as bad as you describe it.
Wish I could take a screen video. Also, plays havoc with JS when trying to reply. At a stage got a few errors. It probably depends on your upstream routing.
 
That makes no sense because the bit that cloudflare does cache does not matter if their CDN is not available. I have never heard of a Cloudflare cache as they serve website traffic from their own CDN and website data is usually dynamic.

Currently Cloudflare is only serving traffic from their local CDN servers for 1.1.1.1 and enterprise customers. So if mybb changes to cloudflare enterprise then traffic will be served locally. I don't think this is up to Seacom at all.
 
We are more than a month later and no sign of this ever getting resolved..... I'm just surprised that Cloudflare is allowing South African traffic to flood the CDNs in Kigali and Djibouti....
 
Top
Sign up to the MyBroadband newsletter
X