Posts Tagged ‘execute’

Rsync copy files with root privileges between servers with root superuser account disabled

Tuesday, December 3rd, 2019

 

rsync-copy-files-between-two-servers-with-root-privileges-with-root-superuser-account-disabled

Sometimes on servers that follow high security standards in companies following PCI Security (Payment Card Data Security) standards it is necessery to have a very weird configurations on servers,to be able to do trivial things such as syncing files between servers with root privileges in a weird manners.This is the case for example if due to security policies you have disabled root user logins via ssh server and you still need to synchronize files in directories such as lets say /etc , /usr/local/etc/ /var/ with root:root user and group belongings.

Disabling root user logins in sshd is controlled by a variable in /etc/ssh/sshd_config that on most default Linux OS
installations is switched on, e.g. 

grep -i permitrootlogin /etc/ssh/sshd_config
PermitRootLogin yes


Many corporations use Vulnerability Scanners such as Qualys are always having in their list of remote server scan for SSH Port 22 to turn have the PermitRootLogin stopped with:

 

PermitRootLogin no


In this article, I'll explain a scenario where we have synchronization between 2 or more servers Server A / Server B, whatever number of servers that have already turned off this value, but still need to
synchronize traditionally owned and allowed to write directories only by root superuser, here is 4 easy steps to acheive it.

 

1. Add rsyncuser to Source Server (Server A) and Destination (Server B)


a. Execute on Src Host:

 

groupadd rsyncuser
useradd -g 1000 -c 'Rsync user to sync files as root src_host' -d /home/rsyncuser -m rsyncuser

 

b. Execute on Dst Host:

 

groupadd rsyncuser
useradd -g 1000 -c 'Rsync user to sync files dst_host' -d /home/rsyncuser -m rsyncuser

 

2. Generate RSA SSH Key pair to be used for passwordless authentication


a. On Src Host
 

su – rsyncuser

ssh-keygen -t rsa -b 4096

 

b. Check .ssh/ generated key pairs and make sure the directory content look like.

 

[rsyncuser@src-host .ssh]$ cd ~/.ssh/;  ls -1

id_rsa
id_rsa.pub
known_hosts


 

3. Copy id_rsa.pub to Destination host server under authorized_keys

 

scp ~/.ssh/id_rsa.pub  rsyncuser@dst-host:~/.ssh/authorized_keys

 

Next fix permissions of authorized_keys file for rsyncuser as anyone who have access to that file (that exists as a user account) on the system
could steal the key and use it to run rsync commands and overwrite remotely files, like overwrite /etc/passwd /etc/shadow files with his custom crafted credentials
and hence hack you 🙂
 

Hence, On Destionation Host Server B fix permissions with:
 

su – rsyncuser; chmod 0600 ~/.ssh/authorized_keys
[rsyncuser@dst-host ~]$

 

For improved security here to restrict rsyncuser to be able to run only specific command such as very specific script instead of being able to run any command it is good to use little known command= option
once creating the authorized_keys

 

4. Test ssh passwordless authentication works correctly


For that Run as a normal ssh from rsyncuser

On Src Host

 

[rsyncuser@src-host ~]$ ssh rsyncuser@dst-host


Perhaps here is time that for those who, think enabling a passwordless authentication is not enough secure and prefer to authorize rsyncuser via a password red from a secured file take a look in my prior article how to login to remote server with password provided from command line as a script argument / Running same commands on many servers 

5. Enable rsync in sudoers to be able to execute as root superuser (copy files as root)

 


For this step you will need to have sudo package installed on the Linux server.

Then, Execute once logged in as root on Destionation Server (Server B)

 

[root@dst-host ~]# grep 'rsyncuser ALL' /etc/sudoers|wc -l || echo ‘rsyncuser ALL=NOPASSWD:/usr/bin/rsync’ >> /etc/sudoers
 

 

Note that using rsync with a ALL=NOPASSWD in /etc/sudoers could pose a high security risk for the system as anyone authorized to run as rsyncuser is able to overwrite and
respectivle nullify important files on Destionation Host Server B and hence easily mess the system, even shell script bugs could produce a mess, thus perhaps a better solution to the problem
to copy files with root privileges with the root account disabled is to rsync as normal user somewhere on Dst_host and use some kind of additional script running on Dst_host via lets say cron job and
will copy gently files on selective basis.

Perhaps, even a better solution would be if instead of granting ALL=NOPASSWD:/usr/bin/rsync in /etc/sudoers is to do ALL=NOPASSWD:/usr/local/bin/some_copy_script.sh
that will get triggered, once the files are copied with a regular rsyncuser acct.

 

6. Test rsync passwordless authentication copy with superuser works


Do some simple copy, lets say copy files on Encrypted tunnel configurations located under some directory in /etc/stunnel on Server A to /etc/stunnel on Server B

The general command to test is like so:
 

rsync -aPz -e 'ssh' '–rsync-path=sudo rsync' /var/log rsyncuser@$dst_host:/root/tmp/


This will copy /var/log files to /root/tmp, you will get a success messages for the copy and the files will be at destination folder if succesful.

 

On Src_Host run:

 

[rsyncuser@src-host ~]$ dst=FQDN-DST-HOST; user=rsyncuser; src_dir=/etc/stunnel; dst_dir=/root/tmp;  rsync -aP -e 'ssh' '–rsync-path=sudo rsync' $src_dir  $rsyncuser@$dst:$dst_dir;

 

7. Copying files with root credentials via script


The simlest file to use to copy a bunch of predefined files  is best to be handled by some shell script, the most simple version of it, could look something like this.
 

#!/bin/bash
# On server1 use something like this
# On server2 dst server
# add in /etc/sudoers
# rsyncuser ALL=NOPASSWD:/usr/bin/rsync

user='rsyncuser';

dst_dir="/root/tmp";
dst_host='$dst_host';
src[1]="/etc/hosts.deny";
src[2]="/etc/sysctl.conf";
src[3]="/etc/samhainrc";
src[4]="/etc/pki/tls/";
src[5]="/usr/local/bin/";

 

for i in $(echo ${src[@]}); do
rsync -aPvz –delete –dry-run -e 'ssh' '–rsync-path=sudo rsync' "$i" $rsyncuser@$dst_host:$dst_dir"$i";
done


In above script as you can see, we define a bunch of files that will be copied in bash array and then run a loop to take each of them and copy to testination dir.
A very sample version of the script rsync_with_superuser-while-root_account_prohibited.sh 
 

Conclusion


Lets do short overview on what we have done here. First Created rsyncuser on SRC Server A and DST Server B, set up the key pair on both copied the keys to make passwordless login possible,
set-up rsync to be able to write as root on Dst_Host / testing all the setup and pinpointing a small script that can be used as a backbone to develop something more complex
to sync backups or keep system configurations identicatial – for example if you have doubts that some user might by mistake change a config etc.
In short it was pointed the security downsides of using rsync NOPASSWD via /etc/sudoers and few ideas given that could be used to work on if you target even higher
PCI standards.

 

Howto debug and remount NFS hangled filesystem on Linux

Monday, August 12th, 2019

nfsnetwork-file-system-architecture-diagram

If you're using actively NFS remote storage attached to your Linux server it is very useful to get the number of dropped NFS connections and in that way to assure you don't have a remote NFS server issues or Network connectivity drops out due to broken network switch a Cisco hub or other network hop device that is routing the traffic from Source Host (SRC) to Destination Host (DST) thus, at perfect case if NFS storage and mounted Linux Network filesystem should be at (0) zero dropped connectios or their number should be low. Firewall connectivity between Source NFS client host and Destination NFS Server and mount should be there (set up fine) as well as proper permissions assigned on the server, as well as the DST NFS should be not experiencing I/O overheads as well as no DNS issues should be present (if NFS is not accessed directly via IP address).
In below article which is mostly for NFS novice admins is described shortly few of the nuances of working with NFS.
 

1. Check nfsstat and portmap for issues

One indicator that everything is fine with a configured NFS mount is the number of dropped NFS connections
or with a very low count of dropped connections, to check them if you happen to administer NFS

nfsstat

 

linux:~# nfsstat -o net
Server packet stats:
packets    udp        tcp        tcpconn
0          0          0          0  


nfsstat is useful if you have to debug why occasionally NFS mounts are getting unresponsive.

As NFS is so dependent upon portmap service for mapping the ports, one other point to check in case of Hanged NFSes is the portmap service whether it did not crashed due to some reason.

 

linux:~# service portmap status
portmap (pid 7428) is running…   [portmap service is started.]

 

linux:~# ps axu|grep -i rpcbind
_rpc       421  0.0  0.0   6824  3568 ?        Ss   10:30   0:00 /sbin/rpcbind -f -w


A useful commands to debug further rcp caused issues are:

On client side:

 

rpcdebug -m nfs -c

 

On server side:

 

rpcdebug -m nfsd -c

 

It might be also useful to check whether remote NFS permissions did not changed with the good old showmount cmd

linux:~# showmount -e rem_nfs_server_host


Also it is useful to check whether /etc/exports file was not modified somehow and whether the NFS did not hanged due to attempt of NFS daemon to reload the new configuration from there, another file to check while debugging is /etc/nfs.conf – are there group / permissions issues as well as the usual /var/log/messages and the kernel log with dmesg command for weird produced NFS client / server or network messages.

nfs-utils disabled serving NFS over UDP in version 2.2.1. Arch core updated to 2.3.1 on 21 Dec 2017 (skipping over 2.2.1.) If UDP stopped working then, add udp=y under [nfsd] in /etc/nfs.conf. Then restart nfs-server.service.

If the remote NFS server is running also Linux it is useful to check its /etc/default/nfs-kernel-server configuration

At some stall cases it might be also useful to remount the NFS (but as there might be a process on the Linux server) trying to read / write data from the remote NFS mounted FS it is a good idea to check (whether a process / service) on the server is not doing I/O operations on the NFS and if such is existing to kill the process in question with fuser
 

linux:~# fuser -k [mounted-filesystem]
 

 

2. Diagnose the problem interactively with htop


    Htop should be your first port of call. The most obvious symptom will be a maxed-out CPU.
    Press F2, and under "Display options", enable "Detailed CPU time". Press F1 for an explanation of the colours used in the CPU bars. In particular, is the CPU spending most of its time responding to IRQs, or in Wait-IO (wio)?
 

3. Get more extensive Mount info with mountstats

 

nfs-utils package contains mountstats command which is very useful in debugging further the issues identified

$ mountstats
Stats for example:/tank mounted on /tank:
  NFS mount options: rw,sync,vers=4.2,rsize=524288,wsize=524288,namlen=255,acregmin=3,acregmax=60,acdirmin=30,acdirmax=60,soft,proto=tcp,port=0,timeo=15,retrans=2,sec=sys,clientaddr=xx.yy.zz.tt,local_lock=none
  NFS server capabilities: caps=0xfbffdf,wtmult=512,dtsize=32768,bsize=0,namlen=255
  NFSv4 capability flags: bm0=0xfdffbfff,bm1=0x40f9be3e,bm2=0x803,acl=0x3,sessions,pnfs=notconfigured
  NFS security flavor: 1  pseudoflavor: 0

 

NFS byte counts:
  applications read 248542089 bytes via read(2)
  applications wrote 0 bytes via write(2)
  applications read 0 bytes via O_DIRECT read(2)
  applications wrote 0 bytes via O_DIRECT write(2)
  client read 171375125 bytes via NFS READ
  client wrote 0 bytes via NFS WRITE

RPC statistics:
  699 RPC requests sent, 699 RPC replies received (0 XIDs not found)
  average backlog queue length: 0

READ:
    338 ops (48%)
    avg bytes sent per op: 216    avg bytes received per op: 507131
    backlog wait: 0.005917     RTT: 548.736686     total execute time: 548.775148 (milliseconds)
GETATTR:
    115 ops (16%)
    avg bytes sent per op: 199    avg bytes received per op: 240
    backlog wait: 0.008696     RTT: 15.756522     total execute time: 15.843478 (milliseconds)
ACCESS:
    93 ops (13%)
    avg bytes sent per op: 203    avg bytes received per op: 168
    backlog wait: 0.010753     RTT: 2.967742     total execute time: 3.032258 (milliseconds)
LOOKUP:
    32 ops (4%)
    avg bytes sent per op: 220    avg bytes received per op: 274
    backlog wait: 0.000000     RTT: 3.906250     total execute time: 3.968750 (milliseconds)
OPEN_NOATTR:
    25 ops (3%)
    avg bytes sent per op: 268    avg bytes received per op: 350
    backlog wait: 0.000000     RTT: 2.320000     total execute time: 2.360000 (milliseconds)
CLOSE:
    24 ops (3%)
    avg bytes sent per op: 224    avg bytes received per op: 176
    backlog wait: 0.000000     RTT: 30.250000     total execute time: 30.291667 (milliseconds)
DELEGRETURN:
    23 ops (3%)
    avg bytes sent per op: 220    avg bytes received per op: 160
    backlog wait: 0.000000     RTT: 6.782609     total execute time: 6.826087 (milliseconds)
READDIR:
    4 ops (0%)
    avg bytes sent per op: 224    avg bytes received per op: 14372
    backlog wait: 0.000000     RTT: 198.000000     total execute time: 198.250000 (milliseconds)
SERVER_CAPS:
    2 ops (0%)
    avg bytes sent per op: 172    avg bytes received per op: 164
    backlog wait: 0.000000     RTT: 1.500000     total execute time: 1.500000 (milliseconds)
FSINFO:
    1 ops (0%)
    avg bytes sent per op: 172    avg bytes received per op: 164
    backlog wait: 0.000000     RTT: 2.000000     total execute time: 2.000000 (milliseconds)
PATHCONF:
    1 ops (0%)
    avg bytes sent per op: 164    avg bytes received per op: 116
    backlog wait: 0.000000     RTT: 1.000000     total execute time: 1.000000 (milliseconds)


nfs-utils disabled serving NFS over UDP in version 2.2.1. Arch core updated to 2.3.1 on 21 Dec 2017 (skipping over 2.2.1.) If UDP stopped working then, add udp=y under [nfsd] in /etc/nfs.conf. Then restart nfs-server.service.
 

4. Check for firewall issues
 

If all fails make sure you don't have any kind of firewall issues. Sometimes firewall changes on remote server or somewhere in the routing servers might lead to stalled NFS mounts.

 

To use properly NFS as you should know as a minimum you need to have opened as ports is Port 111 (TCP and UDP) and 2049 (TCP and UDP) on the NFS server (side) as well as any traffic inspection routers on the road from SRC (Linux client host) and NFS Storage destination DST server.

There are also ports for Cluster and client status (Port 1110 TCP for the former, and 1110 UDP for the latter) as well as a port for the NFS lock manager (Port 4045 TCP and UDP) but having this opened or not depends on how the NFS is configured. You can further determine which ports you need to allow depending on which services are needed cross-gateway.
 

5. How to Remount a Stalled unresponsive NFS filesystem mount

 

At many cases situation with remounting stalled NFS filesystem is not so easy but if you're lucky a standard mount and remount should do the trick.

Most simple way to remout the NFS (once you're sure this might not disrupt any service) – don't blame me if you break something is with:
 

umount -l /mnt/NFS_mnt_point
mount /mnt/NFS_mnt_point


Note that the lazy mount (-l) umount opt is provided here as very often this is the only way to unmount a stalled NFS mount.

Sometimes if you have a lot of NFS mounts and all are inacessible it is useful to remount all NFS mounts, if the remote NFS is responsive this should be possible with a simple for bash loop:

for P in $(mount | awk '/type nfs / {print $3;}'); do echo $P; echo "sudo umount $P && sudo mount $P" && echo "ok :)"; done


If you cd /mnt/NFS_mnt_point and try ls and you get

$ ls
.: Stale File Handle

 

You will need to unmount the FS with forceful mount flag

umount -f /mnt/NFS_mnt_point
 

Sum it up


In this article, I've shown you a few simple ways to debug what is wrong with a Stalled / Hanged NFS filesystem present on a NFS server mounted on a Linux client server.
Above was explained the common issues caused by NFS portmap (rpcbind) dependency, how to its status is fine, some further diagnosis with htop and mountstat was pointed. I've pointed the minimum amount of TCP / UDP ports 2049 and 111 that needs to be opened for the NFS communication to work and finally explained on how to remount a stalled NFS single or all attached mount on a NFS client to restore to normal operations.
As NFS is a whole ocean of things and the number of ways it is used are too extensive this article is just a general info useful for the NFS dummy admin for more robust configs read some good book on NFS such as Managing NFS and NIS, 2nd Edition – O'Reilly Media and for Kernel related NFS debugging make sure you check as a minimum ArchLinux's NFS troubleshooting guide and sourceforge's NFS Troubleshoting and Optimizing NFS Performance guides.

 

Enabling .cgi scripts to execute whenever invoked in Apache

Friday, January 15th, 2010

I wanted to enable .cgi execution on my apache server.
I wanted to set a custom .cgi script to execute as a DirectoryIndex
In order to achieve that I had to add the following to my VirtualHost directive
AddHandler cgi-script .cgi .pl