My video problems mentioned in a previous thread are gone, though I do not know why.
Now my problem is that whenever I have a browser open and an internet connection, my Centos 7 slows to a crawl. Chromium seems to be the least bad. Sometimes it slows to the point that I cannot even move the mouse. Even switching between virtual terminals takes a while sometimes. When I get there, top generally shows me between two and five D states. I've changed service providers lately. The problem survived the change. Some sites seem to aggravate the problem. Most recently, a site tried to connect to c.amazon.adsystem.com . I got the message waiting for c.amazon.adsystem.com . I have that address in my hosts file as 127.0.0.1 . ping finds it and pings. nslookup does not:
[hennebry@localhost ~]$ nslookup c.amazon.adsystem.com Server: 192.168.0.1 Address: 192.168.0.1#53
** server can't find c.amazon.adsystem.com: NXDOMAIN
Whether the slow-to-crawl problem is related to the failure to redirect, I do not know. The former is clearly more important. Any suggestions on how to diagnose it? Any ideas on why chromium is waiting for c.amazon.adsystem.com and how I can fix it?
On Sun, 4 Aug 2019 12:38:29 -0500 (CDT) Michael Hennebry wrote:
Any suggestions on how to diagnose it?
What happens if you try downloading a large file with wget?
What happens if you try browsing some websites with elinks?
On Sun, 4 Aug 2019, Frank Cox wrote:
On Sun, 4 Aug 2019 12:38:29 -0500 (CDT) Michael Hennebry wrote:
Any suggestions on how to diagnose it?
What happens if you try downloading a large file with wget?
I'll try it. My expectation is that it will work just fine *once it starts*. That is my experience with downloading using a browser. In the case of wget, the issue will be typinng the command. Suggestion for a file? A Centos iso perhaps? I'd look for it with a browser, but it might take a while.
What happens if you try browsing some websites with elinks?
What is an elink?
To handle my centos mail, I ssh to the server with my mail and run alpine as the ordinary user that I am. ssh is running from virtual teminal 2 while virtual terminal 1 is crawling.
On Sun, Aug 4, 2019 at 11:16 AM Michael Hennebry < hennebry@web.cs.ndsu.nodak.edu> wrote:
On Sun, 4 Aug 2019, Frank Cox wrote:
What happens if you try browsing some websites with elinks?
What is an elink?
elinks is a text mode browser, as is links, and the venerable lynx
On Aug 4, 2019, at 2:16 PM, Michael Hennebry hennebry@web.cs.ndsu.nodak.edu wrote:
I'll try it. My expectation is that it will work just fine *once it starts*. That is my experience with downloading using a browser. In the case of wget, the issue will be typinng the command. Suggestion for a file? A Centos iso perhaps? I'd look for it with a browser, but it might take a while.
This, with your earlier complaints about resolving Amazon ad service DNS, makes me think it’s DNS-related.
Are you using your ISP’s DNS servers? Are all of them correct? You mentioned you changed ISPs recently, did you switch to the new ISP’s DNS? What happens if you set your DNS to the Quad9[1] DNS server, 9.9.9.9?
-- Jonathan Billings billings@negate.org
elinks does not seem to be working for me. I typed in google.com as my first url. There seems to be no way out of google, nor any way further in. No place to type a url. What appears to be the search window is black and does not accept input. Oops. Now I seem to have clicked on google help or something. There seems no way to back up. Ok. Found the left arrow. Still no way to search or to get out of google (except quit).
How do I adjust my DNS? As noted, my problem has survived a change in ISPs, but they could both be wrong.
On Sun, 4 Aug 2019, Michael Hennebry wrote:
No place to type a url.
Found g.
I'm finding elinks hard to navigate, but at least it's not slowing stuff to a crawl either.
On Sun, 4 Aug 2019, Michael Hennebry wrote:
I'm finding elinks hard to navigate, but at least it's not slowing stuff to a crawl either.
Might have written too soon. elinks is starting to slow down, e.g. down arrow sometimes takes a full minute to respond.
your DNS settings are in /etc/resolv.conf, just like every other unix system since forever.
On Sun, 4 Aug 2019, John Pierce wrote:
your DNS settings are in /etc/resolv.conf, just like every other unix system since forever.
Much to my surprise, I found this: # Generated by NetworkManager search midcoip.net nameserver 192.168.0.1 nameserver 2001:48f8:3004:2ce:5a19:f8ff:fe9e:a4bc
I doubt it's the source of my problems, but the second line looks like something midco did to me somehow. I have noticed a midco search when I wasn't expecting one. My problems predate midco.
Something I just remembered because I saw it again: When I start chromium, I keep getting pop-ups to enter the password to unlock my login keyring. Me no have keyring, except the metal things in my pockets.
Infected Chromium apps are all over the place now. They auto-install and make themselves preferred browsers that auto-start after reboots.
Very bad.
On Aug 4, 2019, at 7:11 PM, Michael Hennebry hennebry@web.cs.ndsu.nodak.edu wrote:
Something I just remembered because I saw it again: When I start chromium, I keep getting pop-ups to enter the password to unlock my login keyring. Me no have keyring, except the metal things in my pockets.
Cheers, Bee
Sounds like you need to go through your packages and uninstall everything not essential!!!
Jay
Something I just remembered because I saw it again: When I start chromium, I keep getting pop-ups to enter the password to unlock my login keyring. Me no have keyring, except the metal things in my pockets.
-- Michael hennebry@web.cs.ndsu.NoDak.edu "Sorry but your password must contain an uppercase letter, a number, a haiku, a gang sign, a heiroglyph, and the blood of a virgin." -- someeecards _______________________________________________ CentOS mailing list CentOS@centos.org https://lists.centos.org/mailman/listinfo/centos
are you running a name server on 192.168.0.1 ? what that ipv6 address ?
On Sun, Aug 4, 2019 at 3:53 PM Michael Hennebry < hennebry@web.cs.ndsu.nodak.edu> wrote:
On Sun, 4 Aug 2019, John Pierce wrote:
your DNS settings are in /etc/resolv.conf, just like every other unix system since forever.
Much to my surprise, I found this: # Generated by NetworkManager search midcoip.net nameserver 192.168.0.1 nameserver 2001:48f8:3004:2ce:5a19:f8ff:fe9e:a4bc
I doubt it's the source of my problems, but the second line looks like something midco did to me somehow. I have noticed a midco search when I wasn't expecting one. My problems predate midco.
-- Michael hennebry@web.cs.ndsu.NoDak.edu "Sorry but your password must contain an uppercase letter, a number, a haiku, a gang sign, a heiroglyph, and the blood of a virgin." -- someeecards _______________________________________________ CentOS mailing list CentOS@centos.org https://lists.centos.org/mailman/listinfo/centos
On Sun, 4 Aug 2019, John Pierce wrote:
are you running a name server on 192.168.0.1 ? what that ipv6 address ?
I expect that that is in the box with midco's router. Do not know about the ipv6 address. I was about to show to what I had changed resolv.conf, but something changed it back. Grrrr. I know I didn't just forget to save it: I tested it with nslookup:
[hennebry@localhost ~]$ nslookup google.com Server: 9.9.9.9 Address: 9.9.9.9#53
Non-authoritative answer: Name: google.com Address: 172.217.5.238
Here is the current resolv.conf:
# Generated by NetworkManager # search midcoip.net nameserver 9.9.9.9 nameserver 192.168.0.1 nameserver 2001:48f8:3004:2ce:5a19:f8ff:fe9e:a4bc
When I point elinks at 192.168.0.1 I get http://192.168.0.1/login.php above a big blank window.
On Sun, Aug 4, 2019 at 7:53 PM Michael Hennebry < hennebry@web.cs.ndsu.nodak.edu> wrote:
On Sun, 4 Aug 2019, John Pierce wrote:
are you running a name server on 192.168.0.1 ? what that ipv6 address ?
I expect that that is in the box with midco's router. Do not know about the ipv6 address. I was about to show to what I had changed resolv.conf, but something changed it back. Grrrr. .... Here is the current resolv.conf:
# Generated by NetworkManager
So you need to modify the source file that NetworkManager is using. somewhere in /etc/network or /etc/networking-scripts, a config file has DNS0=192.168.0.1 or sokmething, or your system is getting that from DHCP
the web login on 192.168.0.1 is undoubtably your modem/router.
On Sun, 4 Aug 2019, John Pierce wrote:
So you need to modify the source file that NetworkManager is using. somewhere in /etc/network or /etc/networking-scripts, a config file has DNS0=192.168.0.1 or sokmething, or your system is getting that from DHCP
Will check on that.
the web login on 192.168.0.1 is undoubtably your modem/router.
On Mon, 5 Aug 2019, Michael Hennebry wrote:
On Sun, 4 Aug 2019, John Pierce wrote:
So you need to modify the source file that NetworkManager is using. somewhere in /etc/network or /etc/networking-scripts, a config file has DNS0=192.168.0.1 or sokmething, or your system is getting that from DHCP
Will check on that.
I've chacked on that. I've made what seemed like promissing changes to /etc/sysconfig/network-scripts/ifup-post and /etc/sysconfig/network-scripts/network-functions . No go. I still get the search line in resolv.conf . I've tried putting in search google.com , but on reboot, it still gives me midco and only midco .
Any idea what does affect search in resolv.conf ? How can I fix this so I do not have to manually edit resolv.conf after each reboot.
On Mon, Dec 02, 2019 at 06:51:44PM -0600, Michael Hennebry wrote:
I've chacked on that. I've made what seemed like promissing changes to /etc/sysconfig/network-scripts/ifup-post and /etc/sysconfig/network-scripts/network-functions . No go. I still get the search line in resolv.conf . I've tried putting in search google.com , but on reboot, it still gives me midco and only midco .
Any idea what does affect search in resolv.conf ? How can I fix this so I do not have to manually edit resolv.conf after each reboot.
Neither of those files are the correct files to edit.
You want to edit the ifcfg-<interface name> file (replace <interface name> with the name of the network interface) and add PEERDNS=no in the file. This will make it so DHCP doesn't overwrite the resolv.conf.
If you just want to hard-code DNS servers, you can either do that in your networkmanager configuration, or add DNS1=9.9.9.9 to the ifcfg file. You can add a second IP with DNS2=1.1.1.1, too. Use whatever DNS IPs you want.
This is documented in /usr/share/doc/initscripts-*/sysconfig.txt
On Tue, 3 Dec 2019, Jonathan Billings wrote:
On Mon, Dec 02, 2019 at 06:51:44PM -0600, Michael Hennebry wrote:
Any idea what does affect search in resolv.conf ? How can I fix this so I do not have to manually edit resolv.conf after each reboot.
Neither of those files are the correct files to edit.
You want to edit the ifcfg-<interface name> file (replace <interface name> with the name of the network interface) and add PEERDNS=no in the file. This will make it so DHCP doesn't overwrite the resolv.conf.
I only have ifcfg-lo , which I am pretty sure is the loopback interface. As expected, other files suggest eth0 is my ethernet connection. Should I add a one-line ifcfg-eth0 file?
If you just want to hard-code DNS servers, you can either do that in your networkmanager configuration, or add DNS1=9.9.9.9 to the ifcfg file. You can add a second IP with DNS2=1.1.1.1, too. Use whatever DNS IPs you want.
What I want is to keep midco from hijacking searches. Keeping midco's DHCP from adding a search line to resolv.conf seems to be the key.
I'll hard-code DNS servers if I have to, but that is not the point of this.
Currently I have # Generated by NetworkManager # search midcoip.net nameserver 2001:48f8:3004:2ce:5a19:f8ff:fe9e:a4bc nameserver 192.168.0.1
My understanding is that the network manager generates resolv.conf and DHCP edits it. Does the network manager include nameservers? Would putting PEERDNS=no in a ifcfg-eth0 file result in an empty resolv.conf ? I realize I could just try it and see, but I do not know how much damage I could do if I mess it up.
This is documented in /usr/share/doc/initscripts-*/sysconfig.txt
I have /usr/share/doc/initscripts-9.49.47/sysconfig.txt only.
On Tue, Dec 03, 2019 at 01:17:53PM -0600, Michael Hennebry wrote:
I only have ifcfg-lo , which I am pretty sure is the loopback interface. As expected, other files suggest eth0 is my ethernet connection. Should I add a one-line ifcfg-eth0 file?
If you are using ethernet (and not a wireless device or some other internet connectivity) you should have an ifcfg- file for the interface. Look at the output of 'ip link' to see all the interfaces you have configured.
Currently I have # Generated by NetworkManager # search midcoip.net nameserver 2001:48f8:3004:2ce:5a19:f8ff:fe9e:a4bc nameserver 192.168.0.1
My understanding is that the network manager generates resolv.conf and DHCP edits it. Does the network manager include nameservers? Would putting PEERDNS=no in a ifcfg-eth0 file result in an empty resolv.conf ? I realize I could just try it and see, but I do not know how much damage I could do if I mess it up.
Since there's no ifcfg file, I suspect you're using NetworkManager to manage your network. Run 'nmcli con' to get a list of your network connections, and then run 'nmcli con edit "Connection Name"' (replace "Connection Name" with the name of your ethernet connection). You can then set ipv4.ignore-auto-dns to 'yes' and then set ipv4.dns and ipv4.dns-search. This should override what DHCP sets.
This is documented in /usr/share/doc/initscripts-*/sysconfig.txt
I have /usr/share/doc/initscripts-9.49.47/sysconfig.txt only.
I used the * as a glob, not a literal character.
On Tue, 3 Dec 2019, Jonathan Billings wrote:
On Tue, Dec 03, 2019 at 01:17:53PM -0600, Michael Hennebry wrote:
I only have ifcfg-lo , which I am pretty sure is the loopback interface. As expected, other files suggest eth0 is my ethernet connection. Should I add a one-line ifcfg-eth0 file?
If you are using ethernet (and not a wireless device or some other internet connectivity) you should have an ifcfg- file for the interface. Look at the output of 'ip link' to see all the interfaces you have configured.
[root@localhost ~]# ip link 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 2: enp0s25: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000 link/ether 00:23:7d:4d:98:89 brd ff:ff:ff:ff:ff:ff [root@localhost ~]#
Currently I have # Generated by NetworkManager # search midcoip.net nameserver 2001:48f8:3004:2ce:5a19:f8ff:fe9e:a4bc nameserver 192.168.0.1
Since there's no ifcfg file, I suspect you're using NetworkManager to manage your network. ...
I'm using the default I got when I installed centos 7. Was not aware of any alternatives.
... Run 'nmcli con' to get a list of your network
[root@localhost ~]# nmcli con NAME UUID TYPE DEVICE Wired connection 1 e4c3a05e-5f33-3ae8-af71-31b70ecf040b ethernet enp0s25 [root@localhost ~]#
connections, and then run 'nmcli con edit "Connection Name"' (replace "Connection Name" with the name of your ethernet connection). You can then set ipv4.ignore-auto-dns to 'yes' and then set ipv4.dns and ipv4.dns-search. This should override what DHCP sets.
[root@localhost ~]# nmcli con edit enp0s25 Error: Unknown connection 'enp0s25'. [root@localhost ~]# nmcli con edit eth0 Error: Unknown connection 'eth0'. [root@localhost ~]# nmcli con edit enp0s25 Error: Unknown connection 'enp0s25'. [root@localhost ~]#
The last was a copy and paste.
Would I still get leases from DHCP? Does DHCP set nameservers? Ideally, I'd like to override just search.
This is documented in /usr/share/doc/initscripts-*/sysconfig.txt
I have /usr/share/doc/initscripts-9.49.47/sysconfig.txt only.
I used the * as a glob, not a literal character.
Sorry about that. I'd thought you meant that there was a whole set of them.
On Tue, Dec 03, 2019 at 02:03:10PM -0600, Michael Hennebry wrote:
I'm using the default I got when I installed centos 7. Was not aware of any alternatives.
... Run 'nmcli con' to get a list of your network
[root@localhost ~]# nmcli con NAME UUID TYPE DEVICE Wired connection 1 e4c3a05e-5f33-3ae8-af71-31b70ecf040b ethernet enp0s25
Your connection's name is "Wired connection 1". The device name is enp0s25.
[root@localhost ~]#
connections, and then run 'nmcli con edit "Connection Name"' (replace "Connection Name" with the name of your ethernet connection). You can then set ipv4.ignore-auto-dns to 'yes' and then set ipv4.dns and ipv4.dns-search. This should override what DHCP sets.
[root@localhost ~]# nmcli con edit enp0s25 Error: Unknown connection 'enp0s25'. [root@localhost ~]# nmcli con edit eth0 Error: Unknown connection 'eth0'. [root@localhost ~]# nmcli con edit enp0s25 Error: Unknown connection 'enp0s25'. [root@localhost ~]#
You need to use the connection name, not the device name.
For whatever reason, the problem I was trying to solve seems to have gone away. I can type in firefox's search box without midco stealing searches. Something changed resolv.conf behind my back. search midcoip.net is there again. I hadn't rebooted or changed firefox's preferences. I'm guessing it was when I turned the connection off and on. To be sure my joy was not tempory, I rebooted. midco still not stealing searches.
I wish I knew what the problem had been and what fixed it.
On Tue, 2019-12-03 at 16:33 -0600, Michael Hennebry wrote:
For whatever reason, the problem I was trying to solve seems to have gone away. I can type in firefox's search box without midco stealing searches. Something changed resolv.conf behind my back. search midcoip.net is there again. I hadn't rebooted or changed firefox's preferences. I'm guessing it was when I turned the connection off and on. To be sure my joy was not tempory, I rebooted. midco still not stealing searches.
I wish I knew what the problem had been and what fixed it.
The search parameter in resolv.conf is not related in any way to the searches in the browser. It just defines the domain to be added to hostnames that you want to resolve. So if you do a name lookup for foo, the resolver would add midcoip.net so you get the ip-address for foo.midcoip.net if foo on its own could not be resolved.
It midcoip wanted to steal your browser searches, it would have to hijack google.com, but that would not work as the browser would not connect as the https certificate wold be invalid. So you need to look at the browser settings to see what is happening there. DNS is most likely not involved. /Louis
On Wed, 4 Dec 2019, Louis Lagendijk wrote:
The search parameter in resolv.conf is not related in any way to the searches in the browser. It just defines the domain to be added to hostnames that you want to resolve. So if you do a name lookup for foo, the resolver would add midcoip.net so you get the ip-address for foo.midcoip.net if foo on its own could not be resolved.
It midcoip wanted to steal your browser searches, it would have to hijack google.com, but that would not work as the browser would not connect as the https certificate wold be invalid. So you need to look at the browser settings to see what is happening there. DNS is most likely not involved.
That all makes sense, especially given recent events. That said, Midco had been stealing searches, and unless something changed it beind my back, my browser has always been set to google.com for searches.
Whatever happened is a mystery to me.
On Aug 4, 2019, at 1:38 PM, Michael Hennebry hennebry@web.cs.ndsu.nodak.edu wrote:
Now my problem is that whenever I have a browser open and an internet connection, my Centos 7 slows to a crawl. Chromium seems to be the least bad. Sometimes it slows to the point that I cannot even move the mouse. Even switching between virtual terminals takes a while sometimes. When I get there, top generally shows me between two and five D states.
Are you sure you don’t have other processes or users running on the system? It only happens when you have a network connection? It might also be swapping heavily, check to see how much RAM you have. Check the output of ‘free’.
Look at the syslogs/journal when you’re logged in (in a terminal, run ’sudo journalctl -xfl’). You will see a lot of stuff printed, but it might give you an idea of what’s going on.
-- Jonathan Billings billings@negate.org
On Sun, 4 Aug 2019, Jonathan Billings wrote:
Are you sure you don?t have other processes or users running on the system? It only happens when you have a network connection? It might also be swapping heavily, check to see how much RAM you have. Check the output of ?free?.
Pretty sure. I rebooted this morning.
top - 17:32:20 up 15:47, 6 users, load average: 1.12, 2.55, 1.56 Tasks: 238 total, 3 running, 235 sleeping, 0 stopped, 0 zombie %Cpu(s): 21.5 us, 3.5 sy, 0.0 ni, 74.1 id, 0.8 wa, 0.0 hi, 0.0 si, 0.0 st KiB Mem : 2020144 total, 80436 free, 1416824 used, 522884 buff/cache KiB Swap: 4883724 total, 3887864 free, 995860 used. 149092 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 27121 hennebry 20 0 1636444 218240 49920 R 29.2 10.8 1:31.01 chromium-b+ 27132 hennebry 20 0 1673932 213652 55576 S 7.3 10.6 1:01.19 chromium-b+ 26614 hennebry 20 0 1871712 104840 41084 S 3.7 5.2 0:26.78 chromium-b+ 27186 hennebry 20 0 1495336 109988 47176 S 3.3 5.4 0:16.55 chromium-b+ 27423 hennebry 20 0 1431484 87076 45928 S 2.3 4.3 0:08.33 chromium-b+ 26647 hennebry 20 0 2685156 188444 27308 S 1.0 9.3 2:19.84 chromium-b+ 5962 hennebry 20 0 3617336 111096 19404 S 0.7 5.5 2:25.24 gnome-shell 27174 hennebry 20 0 1481508 114332 48276 S 0.7 5.7 0:07.88 chromium-b+ 27257 hennebry 20 0 1439148 96572 52020 S 0.7 4.8 0:05.98 chromium-b+ 3483 root 20 0 374444 20432 13732 S 0.3 1.0 0:42.88 X 23824 hennebry 20 0 753956 18272 6488 S 0.3 0.9 0:13.63 gnome-term+ 1 root 20 0 128404 4376 2484 S 0.0 0.2 0:07.65 systemd
Usually when I'm having trouble, there are at least two D's. Of course, 'tain't as crawly as it often gets.
Mem: 2020144 1454904 76140 204764 489100 135004 Swap: 4883724 978480 3905244
Look at the syslogs/journal when you?re logged in (in a terminal, run ?sudo journalctl -xfl?). You will see a lot of stuff printed, but it might give you an idea of what?s going on.
I think this qualifies as interesting. I have rather a lot of it: Aug 04 17:28:20 localhost.localdomain chromium-browser.desktop[26614]: [26647:26728:0804/172820.614816:ERROR:latency_info.cc(149)] Surface::TakeLatencyInfoFromFrame, LatencyInfo vector size 102 is too big.
I found it with google, but all the entries froze chromium.
On 5/08/19 10:42 AM, Michael Hennebry wrote:
Mem: 2020144 1454904 76140 204764 489100 135004 Swap: 4883724 978480 3905244
free -h is generally more readable, but...
It's RAM. You basically have a total of 2G ram on the system, you have less than 500M available and are into swap by nearly 1G, so you're swapping heavily. 2G is enough for a minimal install but browsers such as firefox and chrome can easily use a lot of memory fast and trying to run one on a 2G system while doing an install at the same time will get you swapping and slow the system to a crawl.
Peter
On Aug 5, 2019, at 4:12 AM, Peter peter@pajamian.dhs.org wrote:
On 5/08/19 10:42 AM, Michael Hennebry wrote:
Mem: 2020144 1454904 76140 204764 489100 135004 Swap: 4883724 978480 3905244
free -h is generally more readable, but...
It's RAM. You basically have a total of 2G ram on the system, you have less than 500M available and are into swap by nearly 1G, so you're swapping heavily. 2G is enough for a minimal install but browsers such as firefox and chrome can easily use a lot of memory fast and trying to run one on a 2G system while doing an install at the same time will get you swapping and slow the system to a crawl.
Agreed, 2G of RAM for graphical logins and web browsers is not nearly enough. I was using 4G on a system running C7 and it was unusable, I can’t imagine 2G.
-- Jonathan Billings billings@negate.org
On Mon, 5 Aug 2019, Peter wrote:
On 5/08/19 10:42 AM, Michael Hennebry wrote:
Mem: 2020144 1454904 76140 204764 489100 135004 Swap: 4883724 978480 3905244
free -h is generally more readable, but...
It's RAM. You basically have a total of 2G ram on the system, you have less than 500M available and are into swap by nearly 1G, so you're swapping heavily. 2G is enough for a minimal install but browsers such as firefox and chrome can easily use a lot of memory fast and trying to run one on a 2G system while doing an install at the same time will get you swapping and slow the system to a crawl.
To be clear, by "Centos 7 installation", I meant a PC on which Centos 7 was installed.
In any case, Centos 7 has not always been this slow. Presumably something has changed. I've been living with this for several months, but not forever. I can run compilers and stuff without an internet connection, so I could get some work done.
To get that output, I had free running in a loop and waited for the freeze before copy and pasting.
I wasn't surprised by the result. Occasionally top shows kswap0 (I think) in a D state.
Date: Monday, August 05, 2019 10:44:00 -0500 From: Michael Hennebry hennebry@web.cs.ndsu.nodak.edu
To be clear, by "Centos 7 installation", I meant a PC on which Centos 7 was installed.
In any case, Centos 7 has not always been this slow. Presumably something has changed. I've been living with this for several months, but not forever. I can run compilers and stuff without an internet connection, so I could get some work done.
To get that output, I had free running in a loop and waited for the freeze before copy and pasting.
I wasn't surprised by the result. Occasionally top shows kswap0 (I think) in a D state.
Does the system slow down when you have your internet connection enabled, but aren't explicitly using it (i.e., not using a browser)? If so, look at the netstat output (as root) to see what's going on. You may have some process that runs when the connection is enabled that is taking up system/network resources.
Separately, turn off javascript in the browser you are using and see if that has an effect.
- Richard
Richard wrote:
Date: Monday, August 05, 2019 10:44:00 -0500 From: Michael Hennebry hennebry@web.cs.ndsu.nodak.edu
To be clear, by "Centos 7 installation", I meant a PC on which Centos 7 was installed.
In any case, Centos 7 has not always been this slow. Presumably something has changed. I've been living with this for several months, but not forever. I can run compilers and stuff without an internet connection, so I could get some work done.
To get that output, I had free running in a loop and waited for the freeze before copy and pasting.
I wasn't surprised by the result. Occasionally top shows kswap0 (I think) in a D state.
Does the system slow down when you have your internet connection enabled, but aren't explicitly using it (i.e., not using a browser)? If so, look at the netstat output (as root) to see what's going on. You may have some process that runs when the connection is enabled that is taking up system/network resources.
Separately, turn off javascript in the browser you are using and see if that has an effect.
Javascript - if you're using firefox, install NoScript last week.
mark
Date: Monday, August 05, 2019 13:38:49 -0400 From: mark m.roth@5-cent.us
Richard wrote:
Does the system slow down when you have your internet connection enabled, but aren't explicitly using it (i.e., not using a browser)? If so, look at the netstat output (as root) to see what's going on. You may have some process that runs when the connection is enabled that is taking up system/network resources.
Separately, turn off javascript in the browser you are using and see if that has an effect.
Javascript - if you're using firefox, install NoScript last week.
mark
NoScript selectively blocks javascript, it doesn't turn it off -- which for testing purposes, at least, is the goal.
In ff -- about:config - then enter "javascript" in the search line and set "javascript.enabled" to false.
With chrome this can be done in the advanced settings section, under "site settings".
- Richard
On 6/08/19 6:33 AM, Richard wrote:
Javascript - if you're using firefox, install NoScript last week.
NoScript selectively blocks javascript, it doesn't turn it off -- which for testing purposes, at least, is the goal.
In ff -- about:config - then enter "javascript" in the search line and set "javascript.enabled" to false.
With chrome this can be done in the advanced settings section, under "site settings".
This will cause almost complete breakage of an increasing number of modern websites. The vast majority of sites nowadays absolutely rely on JS and will not run or display correctly without it.
Peter
Peter wrote:
On 6/08/19 6:33 AM, Richard wrote:
Javascript - if you're using firefox, install NoScript last week.
NoScript selectively blocks javascript, it doesn't turn it off -- which for testing purposes, at least, is the goal.
In ff -- about:config - then enter "javascript" in the search line and set "javascript.enabled" to false.
With chrome this can be done in the advanced settings section, under "site settings".
This will cause almost complete breakage of an increasing number of modern websites. The vast majority of sites nowadays absolutely rely on JS and will not run or display correctly without it.
Horse hockey. I read a lot of news, and other stuff. Lessee, slashdot, I enable itself, and fsdn, I think it is, and I'm fine. WaPo is fine.
Just pick and choose. I will admit to being aggravated since google *requires* gstatic to be enabled - I assume that's where they're tracking me.
mark
Richard wrote:
Date: Monday, August 05, 2019 13:38:49 -0400 From: mark m.roth@5-cent.us
Richard wrote:
Does the system slow down when you have your internet connection enabled, but aren't explicitly using it (i.e., not using a browser)? If so, look at the netstat output (as root) to see what's going on. You may have some process that runs when the connection is enabled that is taking up system/network resources.
Separately, turn off javascript in the browser you are using and see if that has an effect.
Javascript - if you're using firefox, install NoScript last week.
NoScript selectively blocks javascript, it doesn't turn it off -- which for testing purposes, at least, is the goal.
In ff -- about:config - then enter "javascript" in the search line and set "javascript.enabled" to false.
With chrome this can be done in the advanced settings section, under "site settings".
Right... but it also prevents the site's javascript from loading 15 other pages, including doubleclick, or gigya (really?) or adserver....
mark
On Mon, 5 Aug 2019, Richard wrote:
Does the system slow down when you have your internet connection enabled, but aren't explicitly using it (i.e., not using a browser)?
The slowdown only happens when the browser is open, but I do not have to be using it.
If so, look at the netstat output (as root) to see what's going on. You may have some process that runs when the connection is enabled that is taking up system/network resources.
Separately, turn off javascript in the browser you are using and see if that has an effect.
On 6/08/19 3:44 AM, Michael Hennebry wrote:
In any case, Centos 7 has not always been this slow. Presumably something has changed.
Websites have gotten more resource-intensive. You've run "yum updates" and now have a newer version of Firefox and/or Chrome. Your browsing habits have changed and you browse with more tabs open now.
Firefox for me has always had some amount of RAM leakage and when I used to run with 4G of RAM I had to restart it frequently (but I browse with a lot of tabs open). I currently have my system maxed at 8G and still have to restart FF every couple of days or so to stop the system from swapping. 2G wouldn't even give me enough RAM to not run a browser on my system, so I can only imagine that your usage to date has been extremely simple.
I have two suggestions for you:
1. Run a lightweight desktop such as XFCE instead of Gnome or KDE.
2. Run out and buy more RAM. Max your system out at 4G or 8G or whatever it will take. You will need it and appreciate it.
Good Luck,
Peter
On Tue, 6 Aug 2019, Peter wrote:
On 6/08/19 3:44 AM, Michael Hennebry wrote:
In any case, Centos 7 has not always been this slow. Presumably something has changed.
Websites have gotten more resource-intensive. You've run "yum updates" and now have a newer version of Firefox and/or Chrome. Your browsing habits have changed and you browse with more tabs open now.
More bloated browsers are not hard to believe in. My habits haven't change much, though. Mostly I use a browser for things I want to read and things I want to download. Maybe that is why I'd been getting along with 2GB.
I have two suggestions for you:
- Run a lightweight desktop such as XFCE instead of Gnome or KDE.
I'll try it.
- Run out and buy more RAM. Max your system out at 4G or 8G or whatever it
will take. You will need it and appreciate it.
Maybe. I open the case with fear and trepidation. The first time I opened a PC case, I zapped my video card installing a disk drive. Under the impressing that memory was the most ststic-sensitive thing in a PC, I had a friend install the DDR2 memory I'd bought. 'Twas frightening to watch: like wathing The Cat in the Hat play with one's grandmother's favorite china. So far as I could tell, he totally ignored the possbility that static could do bad things. It worked and I did not have a heart attack.
Also, what is it with DDR2 prices? When I bought DDR2, DDR3 was the norm and I paid hundreds of dollars for DDR2. Do not remember for how much. Now I suspect DDR4 is the norm and am seeing 8GB of DDR2 for less than $30. Huh? DDR3 isn't much more.
I'll need to do some digging to discover whether my box needs DDR2 or DDR3.DDR3 I doubt it's DDR4.
In article alpine.DEB.2.20.1908051718240.19025@mail.cs.ndsu.nodak.edu, Michael Hennebry hennebry@web.cs.ndsu.nodak.edu wrote:
I'll need to do some digging to discover whether my box needs DDR2 or DDR3.DDR3 I doubt it's DDR4.
Do:
# dmidecode | less
and look for the entries for the existing RAM you have. It will also tell you if you have any unpopulated RAM slots ("No module installed").
It won't tell you the maximum size RAM each slot will take. For that, you would need to look up the specs for the motherboard or system (you can find the model number in the dmidecode output too).
Or you can go to the website for a memory vendor such as Crucial or Kingston and enter your model number, and it will tell you what RAM is compatible and what it costs.
Cheers Tony
-----Original Message----- From: Tony Mountifield tony@softins.co.uk Sent: Tuesday, August 06, 2019 5:44 AM To: centos@centos.org Subject: Re: [CentOS] browsers slowing Centos 7 installation to a crawl
In article alpine.DEB.2.20.1908051718240.19025@mail.cs.ndsu.nodak.edu, Michael Hennebry hennebry@web.cs.ndsu.nodak.edu wrote:
I'll need to do some digging to discover whether my box needs DDR2 or
DDR3.DDR3
I doubt it's DDR4.
Do:
# dmidecode | less
and look for the entries for the existing RAM you have. It will also tell you if you have any unpopulated RAM slots ("No module installed").
<SNIP>
In the mean time you could enable zswap[1] to make the most out of the ram and swap you do have. according to [2] either add 'zswap.enabled=1' to the boot line or 'echo 1 > /sys/module/zswap/parameters/enabled'
On an EL6 system with 512M of ram (yes that is correct, .5GB, actually ~.4GB because some is shared with the intel video) I use up-to-date EL6 firefox reasonably comfortably at home using zswap, setting max_pool_percent to 40.
I am trying to remember which kernel 7 is running (I tend to use the elrepo LT kernel on EL6) and if it has zswap or only zram.
before switching to the LT kernel I used zram [3] on the machine as swap space (pretending to be 90% of ram) instead and it works very fast, but when it runs out (say on a site with lots of JPGs) and you hit real swap again performance tanks VERY badly. EL7 may have zram setup to swap in such a way that you can 'sysctl start zram' and try it out.
Good luck.
[1] https://en.wikipedia.org/wiki/Zswap [2] https://www.kernel.org/doc/html/latest/vm/zswap.html [3] https://en.wikipedia.org/wiki/Zram
-- Even when this disclaimer is not here: I am not a contracting officer. I do not have authority to make or modify the terms of any contract.
On Tue, 6 Aug 2019, Peter wrote:
- Run out and buy more RAM. Max your system out at 4G or 8G or whatever it
will take. You will need it and appreciate it.
My fears and trepidations have been realized.
I finally got around to trying to install the memory I bought. No go. The first card seems like it's in almost ok, but will not go far enough down to be latched. The notches seem correct. I cannot even replace the memory I removed. Grrr. The net result seem to be that I destroyed my computer.
Any thoughts on how to undestroy my computer?
On Wed, 9 Oct 2019 17:14:12 -0500 (CDT) Michael Hennebry wrote:
I cannot even replace the memory I removed.
There might be dirt plugging up the slot. Try vacuuming the slot out (carefully) and see if it fits after that.
On Wed, 9 Oct 2019, Frank Cox wrote:
On Wed, 9 Oct 2019 17:14:12 -0500 (CDT) Michael Hennebry wrote:
I cannot even replace the memory I removed.
There might be dirt plugging up the slot. Try vacuuming the slot out (carefully) and see if it fits after that.
The process gave me a better look at the mechanics of latching. Apparently the latches are supposed to activate as one pushes the cards down. Pushing in the middle did not give me enough leverage to activate stiff latches. Once I realized that, pushing on the ends did the trick
It boots and notices 8GB.
In any case, Centos 7 has not always been this slow. Presumably something has changed.
Firefox especially, and to some extent Chrome, have both started using much more memory recently (as in the last six months or so). I run 50+ desktops on CentOS and I've noticed more and more of them getting low on memory more often and almost always Firefox is the culprit. These are systems with 8Gb that are struggling.
P.