Khaled's questions

I am using glusterfs on several Ubuntu servers: two of them are running glusterfs servers in replication mode.

Without any clear error, the glusterfs partition became stale and the system shows this error when I try to access the stale partition:

Transport endpoint is not connected

Also, when running ls -l on the parent folder I get:

d????????? ? ?      ?          ?                ? myfolder

I tried all types of commands that I can find to umount this partition, but I could not get it done:

umount -l /path/to/mount/point
umount -f /path/to/mount/point

Also, using fuser command to show processes accessing this folder did not work. Unload the fuse kernel module can not be done as it is clear from the kernel config that fuse is built into the kernel and not a loadable module. I found this line in /boot/config-2.6.32-24-server

CONFIG_FUSE_FS=y

I have been left with two options:

1. Reboot the system.
2. Create another mount point like myfolder2 and mount this again using sudo glusterfs -f /etc/glustefs/glusterfs.vol /path/to/folder2.

Of course, I have chosen to go with option 2.

Anyone faced such an issue before?

Anyone has a better solution for such a case?

Iptables can be optimized by putting the mostly used rule on the top such as the known related rule that matches after connection establishment. Also, optimization can be done by using jumps to avoid very long chains. This link shows an example.

My question is about optimizing the rule itself. How the performance will be by adding and/or deleting some checks in a specific rule? What about the order of these checks? For example, this rule:

iptables -A FORWARD -i eth0 -s source_ip -d dest_ip -p tcp --dport 80 -j ACCEPT

can be rewritten as:

iptables -A FORWARD -s source_ip -d dest_ip -p tcp --dport 80 -j ACCEPT

Both rules will allow http traffic from specific source to specific destination. Do you think there will be a performance difference between the two? Also, the checks can be re-ordered as:

iptables -A FORWARD -p tcp --dport 80 -s source_ip -d dest_ip -i eth0 -j ACCEPT

Will this also make a difference or iptables will automatically take care of it.

My question in short: "Which is better? To assign private IP addresses to public servers and do NATing, or to assign public IP addresses to them without the need for NATing"?

The NATing can be useful for users to allow many users to access the Internet using a small pool of public IP addresses (can be just one IP). This is called PATing also.

I am discussing NATing the IP addresses of public servers such as: DNS, Web servers, Email servers, etc. NATing of public servers should be done using static NAT. So, someone can think of eliminating the NATing process and assigning the public IPs on the servers. Here is a comparison of the two scenarios.

With NATing:

+ The public IP addresses are managed and used on the gateway as desrired.

+ The public IPs are preserved and can be used as efficiently as possible.

- A little overhead will be added on the gateway router / firewall.

- The private IP addresses may be exposed to the outside (for example, in emails headers, or in a misconfiguration).

- We need to create two views for the DNS servers (private for internal users and public for external users).

Without NATing:

- Assigning more than one IP address for a server should be done on the server itself. The application/service should be aware to use all the IPs for the outgoing traffic.

- The public IP addresses are assigned to the servers directly. So, some IPs may be wasted (not used) such as the case of cluster and VIP.

+ The routing/forwarding will be a bit faster as we are not doing NATing (less overhead).

+ All communications will be done using the public IPs.

+ One DNS view will be enough as we are using the same IPs from internal and external.

Any other ideas/suggestions??

I have ubuntu server 10.04 installed on a Raid10 array (MD) using 4 HD drives.

As it is known, Raid10 is Raid 1 + Raid 0. So, two HD drives are stripped and they are mirrored (or the other way around).

Is there an easy way to figure out which two of these four drives are stripped and which ones are mirrored?

Here is the output of: /proc/mdstat

Personalities : [raid10] [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4]
md0 : active raid10 sda1[0] sdb1[1] sdd1[3] sdc1[2]
      388992 blocks 64K chunks 2 near-copies [4/4] [UUUU]

md2 : active raid10 sda7[0] sdb7[1] sdd7[3] sdc7[2]
      19529600 blocks 64K chunks 2 near-copies [4/4] [UUUU]

md4 : active raid10 sda9[0] sdb9[1] sdd9[3] sdc9[2]
      9762688 blocks 64K chunks 2 near-copies [4/4] [UUUU]

md1 : active raid10 sda6[0] sdb6[1] sdd6[3] sdc6[2]
      19529600 blocks 64K chunks 2 near-copies [4/4] [UUUU]

md5 : active raid10 sda10[0] sdb10[1] sdd10[3] sdc10[2]
      195309440 blocks 64K chunks 2 near-copies [4/4] [UUUU]

md6 : active raid10 sda11[0] sdb11[1] sdd11[3] sdc11[2]
      1558599552 blocks 64K chunks 2 near-copies [4/4] [UUUU]

md3 : active raid10 sda8[0] sdb8[1] sdd8[3] sdc8[2]
      146483072 blocks 64K chunks 2 near-copies [4/4] [UUUU]

unused devices: <none>

I am wondering whether the raid rebuild process will impact the system performance significantly or not. Will the server be able to respond with a reasonable delay during the rebuild process?

Another question came to my mind. Is there any way we can give priority to the disk rebuild process? For example, can we limit the rate at which the data is read/written? Of course, this may be risky because we may have another failure during the rebuild. Let us focus on the performance aspect and ignore this risk!

I am talking about Linux raid mdadm.

I have a machine with two HD drives. I want to use Raid1 for redundancy to avoid data loss.

I have two possible scenarios:

1- Install /boot partition on both drives using Raid1.

2- Install 2 separate /boot partitions on each drive without Raid.

My concern is to be able to the system even if one HD failed. Will this requirement be met in both cases or at least in one them?

I think the performance will not be an issue here.

I am talking about a Linux machine.

Do you know/recommend any tool for graphing traffic based on IP or host? In this case, I will get a graph for every IP that passes through my gateway. Of course, I will use a predefined list of source IP addresses. This is useful to monitor the traffic usage of all hosts inside your network.

I am using nagios grapher to graph the network traffic, but this tool graph the total traffic passing through the system interface(s). Nagios uses a script to monitor the traffic of the interface. After that, the traffic data is passed to nagios grapher.

I need to install such a tool on my Linux server/gateway.

I have a Java application that depends on date/time of the system to check for new and old email messages.

While I was checking the Java application log, I found that the server timezone has changed. This caused the application to fail in doing its job.

What could be the reason of this sudden change in the server timezone? How can I prevent such problems in the future.

I am using ubuntu server 8.04.

It is expected that NFS will be slower than any "normal" file system such as ext3. Transferring files over the network is the main reason for the additional delay.

I am wondering how NFS can be compared with other file systems. Let us assume the NFS server and the clients are located on the same Gigabit LAN. Also, let us assume we have only 10 clients.

Is there any benchmarks that we can look at? Can we say some like the following?

NFS delay ranges from 150% to 200% when compared to other file systems?

Of course, many factors can affect the delay, such as read/write operations on disks and files' sizes. Can we have an approximate measurement of the delay difference?

There are many file systems for a Linux server administrator to choose from. Also, it is not always an easy task to figure out the appropriate layout and optimal file system for your needs.

Can you suggest some guidelines for choosing the "optimal" file system for a Linux server? I know this is a general question. Let us be more specific. I will put it this way.

1- I have a caching web proxy server. The proxy usually needs to store many small objects and also there are some medium and large objects. I think the web server case will be similar.

2- I have an email server. It needs to save the users' messages.

3- I have a database server.

All these three servers require very much IO access (read and write).

For other types of servers, I think it is not so critical. For example, a Linux-based firewall does not require HD access like a proxy or web server. Most of its processing is done in the memory.

I think the best answer will suggest the appropriate file system for each server (preferably with justifications). Let us focus on performance as the factor to base our suggestions on.

If I am using NTP daemon or ntpdate command, do I need to worry about changing the timezone?

Should I re-configure the timezone using cron job to guarantee the accuracy of the server time?

I am using ubuntu server.

Is it possible to use GlusterFS to have more flexible FS and to support higher capacity + RAID-1 to have redundancy to avoid data loss in case of failure?

I need to get a mirrored FS that has high flexibility in scaling (adding more HDs and/or adding more machines).