Folks
I've been having lots of problems with Centos 7 on Raspberry Pi 2. The network seems to work for a while after a bootstrap, but then dies. The command systemctl restart network always fails. This occurs after installation and updates.
I wonder of others have had this issue, or is my RPI2 defective? It's connected to my internal network by a wire-connection to a working Centos 6 gateway. I posted the results of journalctl earlier, so won't bother reposting.
Is this the right mail=list to detect if and when the issue is addressed?
David
On 12/28/2015 10:01 PM, david wrote:
Folks
I've been having lots of problems with Centos 7 on Raspberry Pi 2. The network seems to work for a while after a bootstrap, but then dies. The command systemctl restart network always fails. This occurs after installation and updates.
I wonder of others have had this issue, or is my RPI2 defective? It's connected to my internal network by a wire-connection to a working Centos 6 gateway. I posted the results of journalctl earlier, so won't bother reposting.
No problems on my Cubieboard2: medon.htt-consult.com that has been up 4 1/2 days. The only messages on Network that I see are:
messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> nameserver '2001:558:feed::2' messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> nameserver '2001:558:feed::1' messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 state changed unknown -> bound messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> Policy set 'eth0' (eth0) as default for IPv6 routing and DNS. messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 client pid 1395 exited with status 0 messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 state changed bound -> done
which was when I booted it up in my production rack.
Is this the right mail=list to detect if and when the issue is addressed?
Hopefully others with RPi2 will chime in and it will be here.
At 08:40 PM 12/28/2015, you wrote:
On 12/28/2015 10:01 PM, david wrote:
Folks
I've been having lots of problems with Centos 7 on Raspberry Pi 2. The network seems to work for a while after a bootstrap, but then dies. The command systemctl restart network always fails. This occurs after installation and updates.
I wonder of others have had this issue, or is my RPI2 defective? It's connected to my internal network by a wire-connection to a working Centos 6 gateway. I posted the results of journalctl earlier, so won't bother reposting.
No problems on my Cubieboard2: medon.htt-consult.com that has been up 4 1/2 days. The only messages on Network that I see are:
messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> nameserver '2001:558:feed::2' messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> nameserver '2001:558:feed::1' messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 state changed unknown -> bound messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> Policy set 'eth0' (eth0) as default for IPv6 routing and DNS. messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 client pid 1395 exited with status 0 messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 state changed bound -> done
which was when I booted it up in my production rack.
Is this the right mail=list to detect if and when the issue is addressed?
Hopefully others with RPi2 will chime in and it will be here.
Robert, and others
In one of those Eureka moments, I believe I understand why my network is failing. Examination of the "journalctl" output led me to the following: - Chronyd is a REPLACEMENT for ntp - Running both of them at the same time leads to ntpd disabling the ethernet link.
If I'm right, perhaps a note should be added into the Arm web page so other dummies like me don't make the same mistake.
For extra bonus points, I wonder if someone could answer the question: - Is it OK to use chrony instead of ntp on Centos 6&7 on i386/x86_64 systems? Should I change over?
David
On 12/31/2015 06:13 PM, david wrote:
At 08:40 PM 12/28/2015, you wrote:
On 12/28/2015 10:01 PM, david wrote:
Folks
I've been having lots of problems with Centos 7 on Raspberry Pi 2. The network seems to work for a while after a bootstrap, but then dies. The command systemctl restart network always fails. This occurs after installation and updates.
I wonder of others have had this issue, or is my RPI2 defective? It's connected to my internal network by a wire-connection to a working Centos 6 gateway. I posted the results of journalctl earlier, so won't bother reposting.
No problems on my Cubieboard2: medon.htt-consult.com that has been up 4 1/2 days. The only messages on Network that I see are:
messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> nameserver '2001:558:feed::2' messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> nameserver '2001:558:feed::1' messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 state changed unknown -> bound messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> Policy set 'eth0' (eth0) as default for IPv6 routing and DNS. messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 client pid 1395 exited with status 0 messages-20151227:Dec 24 09:46:33 medon NetworkManager[523]: <info> (eth0): DHCPv6 state changed bound -> done
which was when I booted it up in my production rack.
Is this the right mail=list to detect if and when the issue is addressed?
Hopefully others with RPi2 will chime in and it will be here.
Robert, and others
In one of those Eureka moments, I believe I understand why my network is failing. Examination of the "journalctl" output led me to the following:
- Chronyd is a REPLACEMENT for ntp
- Running both of them at the same time leads to ntpd disabling the
ethernet link.
If I'm right, perhaps a note should be added into the Arm web page so other dummies like me don't make the same mistake.
For extra bonus points, I wonder if someone could answer the question:
- Is it OK to use chrony instead of ntp on Centos 6&7 on i386/x86_64
systems? Should I change over?
I am leaving my C6 (really RSEL6) servers as is with NTP.
all my C7 are using Chrony.
On 1 ianuarie 2016 01:13:22 EET, david david@daku.org wrote:
In one of those Eureka moments, I believe I understand why my network is failing. Examination of the "journalctl" output led me to the following:
- Chronyd is a REPLACEMENT for ntp
Right.
- Running both of them at the same time leads to ntpd disabling the
ethernet link.
If this really happens, it looks like a bug to me. Neither of the two should modify the status of the link.
If I'm right, perhaps a note should be added into the Arm web page so other dummies like me don't make the same mistake.
IMNSHO, if this behaviour is confirmed , it should be filed as a bug and fixed. Network status should under no circumstance depend on a time daemon, whatever that daemon would be.
For extra bonus points, I wonder if someone could answer the question:
- Is it OK to use chrony instead of ntp on Centos 6&7 on i386/x86_64
systems? Should I change over?
ntp (+ ntpdate) is standard in EL6. AFAIK it has been replaced by chrony in EL7. However chrony is also available (via EPEL if I am not mistaken -- I cannot verify now) for EL6.
At 03:33 PM 12/31/2015, you wrote:
On 1 ianuarie 2016 01:13:22 EET, david david@daku.org wrote:
In one of those Eureka moments, I believe I understand why my network is failing. Examination of the "journalctl" output led me to the following:
- Chronyd is a REPLACEMENT for ntp
Right.
- Running both of them at the same time leads to ntpd disabling the
ethernet link.
If this really happens, it looks like a bug to me. Neither of the two should modify the status of the link.
If I'm right, perhaps a note should be added into the Arm web page so other dummies like me don't make the same mistake.
IMNSHO, if this behaviour is confirmed , it should be filed as a bug and fixed. Network status should under no circumstance depend on a time daemon, whatever that daemon would be.
For extra bonus points, I wonder if someone could answer the question:
- Is it OK to use chrony instead of ntp on Centos 6&7 on i386/x86_64
systems? Should I change over?
ntp (+ ntpdate) is standard in EL6. AFAIK it has been replaced by chrony in EL7. However chrony is also available (via EPEL if I am not mistaken -- I cannot verify now) for EL6.
New Years Eve Update Having eliminated NTP from my configuration, survival time went from 20 mins to almost an hour. But I'm still getting network fails, and nothing in journalctl or /var/log/messages seems to point to the failure. Of course, "systemctl restart network" hard fails, and the only way to restore network connectivity is a reboot.
Does anyone have a clue where I should look? I can isolate the failure to a 5-second window (I have another machine probing it every five seconds), but see nothing in the aforementioned logs. Any cluee? anyone?
David
On 1 ianuarie 2016 08:47:02 EET, david david@daku.org wrote:
But I'm still getting network fails, and nothing in journalctl or /var/log/messages seems to point to the failure. Of course, "systemctl restart network" hard fails,
could you please tetry with "systemctl restart NetworkManager" ? The network service is not normally used in EL7 ,it has been supetseded by NM.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 29/12/15 04:01, david wrote:
Folks
I've been having lots of problems with Centos 7 on Raspberry Pi 2. The network seems to work for a while after a bootstrap, but then dies. The command systemctl restart network always fails. This occurs after installation and updates.
I wonder of others have had this issue, or is my RPI2 defective? It's connected to my internal network by a wire-connection to a working Centos 6 gateway. I posted the results of journalctl earlier, so won't bother reposting.
Is this the right mail=list to detect if and when the issue is addressed?
David
Hmm, multiple people reported the same image running fine on the rpi2. Does that happen only after some intense network activity ? What does journalctl / /var/log/messages have to say about the issue ? Also, the default image doesn't use network , but NetworkManager, so systemctl status NetworkManager is the way to go, and not systemctl status network
The network issue you seems to report rings a bell on my side, but not with the rpi2 board, and not with the centos 7 image. I had that issue with RedSleeve 6 running on my rpi1, and the fix was to change some default kernel parameter for that issue. Here is my ansible playbook snippet for that :
- name: configuring kernel memory for network issue sysctl: name=vm.min_free_kbytes value=8192 state=present
But that was for my rpi1 board, so don't think that it applies to rpi2 and kernel 4.1.11-v7+.
Can you create a bug report on https://bugs.centos.org, so that we can track it there, and so that it can also appear publicly, in case other people would have the same issue (my rpi2 board is running for quite some weeks now without a glitch, but haven't really transferred a lot of data through the ethernet interface either)
Kind Regards,
- -- Fabian Arrotin The CentOS Project | http://www.centos.org gpg key: 56BEC54E | twitter: @arrfab
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 29/12/15 08:23, Fabian Arrotin wrote:
On 29/12/15 04:01, david wrote:
Folks
I've been having lots of problems with Centos 7 on Raspberry Pi 2. The network seems to work for a while after a bootstrap, but then dies. The command systemctl restart network always fails. This occurs after installation and updates.
I wonder of others have had this issue, or is my RPI2 defective? It's connected to my internal network by a wire-connection to a working Centos 6 gateway. I posted the results of journalctl earlier, so won't bother reposting.
Is this the right mail=list to detect if and when the issue is addressed?
David
Hmm, multiple people reported the same image running fine on the rpi2. Does that happen only after some intense network activity ? What does journalctl / /var/log/messages have to say about the issue ? Also, the default image doesn't use network , but NetworkManager, so systemctl status NetworkManager is the way to go, and not systemctl status network
The network issue you seems to report rings a bell on my side, but not with the rpi2 board, and not with the centos 7 image. I had that issue with RedSleeve 6 running on my rpi1, and the fix was to change some default kernel parameter for that issue. Here is my ansible playbook snippet for that :
- name: configuring kernel memory for network issue sysctl:
name=vm.min_free_kbytes value=8192 state=present
But that was for my rpi1 board, so don't think that it applies to rpi2 and kernel 4.1.11-v7+.
Can you create a bug report on https://bugs.centos.org, so that we can track it there, and so that it can also appear publicly, in case other people would have the same issue (my rpi2 board is running for quite some weeks now without a glitch, but haven't really transferred a lot of data through the ethernet interface either)
Kind Regards,
So, just to add that , without any change to the the rpi2 CentOS 7 userland image, I've downloaded 100 times the CentOS-7-i386-LiveGNOME-1511.iso image (so 1.1GB * 100) on the rpi2, from an internal mirror (rpi2 connected on a gbit switch), and no network issue, even under high load.
I must even admit that the throughput I was able to get impressed me (as the ethernet interface is just a smsc95xx USB 2.0 Ethernet one): 2015-12-29 11:38:18 (11.2 MB/s)
It would be interesting to have other informations about: - - logs - - which kind of traffic is that rpi2 getting (push/pull/both)
- -- Fabian Arrotin The CentOS Project | http://www.centos.org gpg key: 56BEC54E | twitter: @arrfab
-
david -
Fabian Arrotin -
Manuel Wolfshant -
Robert Moskowitz