|My last post was years ago. I reach limit of disks in server chassis. I've 19 2TB disks and storage is almost full. I've decided to migrate to newer RAID card becouse old ones don't support bigger disks than 2TB. At the begiining I'll use two 8TB disks. Now data is moved to new volume. When it finished I'll remove one old RAID card and bunch of 2TB disks to make room for new ones.
|I've nasty problem with root filesystem in ftp server. Both of disks in RAID1 got bad sectors. Without notification server just hanged. I've replaced disks and I hope now it will be all right.
|To find out what is using port on linux you can use netstat -tulpn
|I had 1.9TB xfs volume. I use xfs_growfs to expand it. After this I got 'No space left on device' error message. Solution is to unmount volume and mount it with -inode64 options. Or remount if you wish.
|When your apache doesn't start and you see in error_log 'No space on device' but it's not true becouse you have free space. Probably you have problem with semaphores. Solution : remove all apache semaphores|
for i in `ipcs -s | grep [your apache user] | cut -d' ' -f2` ;do ipcrm -s $i;done
|How to extend RAID5 on 3ware with LVM and XFS on nobooting storage.|
1. In 3DM2, choose Management > Maintenance
2. Select the unit you wish to expand by checking the box next to the Unit ID
3. Click the Migrate Unit button
4. The Migrate dialog box appears, listing the drives which can be added to the unit.
5. Select the drive(s) you wish to add to the unit by checking the Port ID box next to each one.
6. If desired or necessary, select RAID level and click OK.
7. After migration process is finished unmount the file system stored on the extended unit.
8. Remove information about your VG by vgchange -an [your_vg_name] .
9. Update system about new unit size. In 3DM2 choose Management > Maintenance , select the appropriate unit and click Remove Unit button.
10. After the unit has been removed, click the Rescan button. Wait couple of seconds and new device will be detected by system.
11. Use parted to correct GPT table: parted /dev/sdj and choose Fix one or two times.
12. Create new partition gdisk /dev/sdj (sdj is my device replace it with your name)
13. Type n, next Enter, next Enter, type w, next Enter, type Y, Enter.
14. Activate your VG by vgchange -ay [your_vg_name]
15. Mount your LV by mount /dev/[your LV name]
16. Create new Physical Volume by pvcreate /dev/sdj2 .
17. Extend Volume Group by vgextend [your_vg_name] /dev/sdj2
18. Extend your Logical Volume by lvextend /dev/[your LV location] /dev/sdj2
19. Extend XFS online by xfs_growfs -d [your mount point]
20. Be happy with your extended volume.
|When you attach disk with created filesystem to the 3ware 9550 controller and you will create 'single' disk based on it. You will lost your data. I read that the reverse operation apparently doesn't cause lost of data. We will see
|One month ago I put into server HDD with HPA enable. 2TB disk had incorrect size.I have to deal with it. First|
root@foopc:~# hdparm -N /dev/sdb
max sectors = 1844674408/3907029168, HPA is enabled
root@foopc:~# hdparm -N p3907029168
max sectors = 18446744073321613488/3907029168, HPA setting seems invalid
root@foopc:~# echo x > /sys/bus/scsi/devices/5:0:0:0/delete
Wait couple of seconds. Drive will spin down
root@foopc:~# echo "- - -" >/sys/class/scsi_host/host1/scan
Drive is up with proper size. I saw in logs
ata6: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
ata6.00: ATA-8: WDC WD20EARS-00MVWB0, 51.0AB51, max UDMA/133
ata6.00: 3907029168 sectors, multi 16: LBA48 NCQ (depth 0/32)
ata6.00: configured for UDMA/100
ata6: EH complete
sd 5:0:0:0: Attached scsi disk sdb
sd 5:0:0:0: Attached scsi generic sg1 type 0
|During data moving I realize that two of hard disks are broken. One is totaly dead. I have to improve monitoring system to better discovering such situations. I lost Fedora archive volumen. Fedora archive has slow server. Rsync is working for almost two weeks and still doesn't finish. Fedora archive rpms are added to database again.
|I've added 3 x 2TB harddisks to ftp server. I created RAID5 and I'm moving data from smallest disks on raid
|I redesigned database. I decided to remove possibility to search across changelog, requires, summary and description. I saw that such search was not popular or even not used. Thanks to that you get faster search results. It's important becouse there is over 3.5 mln rpms in database now. I'm looking for cheap HP DL365 G1 server to build HA cluster to improve stability.
|It was a bad Saturday. I had to add new disk to database server. After that server reported that it is broken. I see FF error on motherboard display. It means that neither motherboard or cpu or psu is broken. Great. It take me couple of hours to bring back server to life. I clean and check everything. It seems that now it's all right. It is a great thing to have backup database server. :-)
|I added new functionality to pbone. Now MAN pages are displayed on pbone. Data is uploading to database. So every second new man pages are added to database. There is plenty of information. It seems that my current SCSI disks are too small. I have to replace current disks by bigger ones.
|I'd like to make some update. I wanted to add bigger disk to pbone ftp but did not succeed. It's hard to make changes in case where I have 14 hard disks. I accidentally turned off other disk and locked one processors fan. It take me about hour to discover couse my problems with hardware. I thought that PSU in ftp server is broken becouse server was shut down after a few minutes after turn on. If you have a lot of disks and you wait couple of minutes when you see 'Mounting local filesystems:'. I suggest to add -v option to mount in rc.sysinit file. After that you will see which disk is mounted
|Today pbone was offline couple of hours. Suddenly high end 1200W PSU just switch off. It's broken. My old good SCSI PCI-X adapter is also cold. I had to come back to onboard SCSI adapter. For now it's working. I had to remove couple RAM sticks to lower power consuption. After that I could use old 560W PSU. Search engine will run a little slower.
|So it seems that I have some kind of stable situation with hardware. It looks like onboard SCSI adapter in database server is unstable. I had to come back to old good SCSI PCI-X adapter. I'm still waiting for replacement for failed SCSI disk. I hope I will get it in two days. It seems that PSU in FTP server is rather overloaded , 2 processors, 8 RAM sticks, 14 hard disks. Maybe it is too much for it. I had to remove smaller HDDs and put into one or two bigger.
|Last week was very bad for pbone. I lost two SCSI hard disks and 6GB RAM from database server. Database server hanged couple times. Database was corrupted. Recovery process took long time. Now I have to buy some hardware.
|I add 4GB RAM to database server and 4GB RAM to ftp server. DB server has 30GB RAM. It's nice number. Thanks to last donation I'll be able to add bigger SCSI disks to DB server. Yum repository is working well IMHO. I wonder if I should merge noarch repo with i386 and x86_64 repos. Database files have 43GB size. Is it already big database or not ?
|I add 4GB RAM to database server. Now it has 26GB RAM. I change PSU in database server. Now I have 1200W PSU.
|I add next 2 x 2GB RAM modules to database server. I've to reorganize connection between database server and web server. Now here I've bottleneck
|I just add 2 x 2GB RAM modules to database server. I hope you will see better performance but it isn't the end of needs.
|Since three weeks ago I see more visitors. In July PBONE has more then 1.5 milion visitors. That's why you see that server slow down again. Last upgrade was necessary but I see that next upgrade is the must. I need more RAM and bigger SCSI hard disks. Now database takes about 30GB space . Only more powerful server is able to handle such big database and process queries in reasonable time. I count on yours donations. Server really need it. Maybe you work in company which want to be PBONE official sponsor and has small banner on every page for one year. IMHO 2000$ is not so much for a company.
|I've upgraded database server. Now it is quad Opteron dual core. I see speed up. From now I hope information about 'database connection lost' will disappear. Of course uprade was stresful. First, mobo for quad Opterons is big, very big. First in my life I see W-ATX (13'-16'). All my big tower chassis are not designed to carry such mobo. So now I don't use any chassis. This is unsafe but I haven't choice. Second, I have problem with onboard SCSI controller. Kernel panic during booting due to unable to read from unknown device. Strange becouse I have support for all LSI stuff in kernel. I have to use extra Adaptec SCSI controller. It works.
|I've started replication mysql database. All updates go to master. All queries go to slave. Now I know that info about 50000 rpms are waiting for insert. I have to completely redesign database or buy faster hardware. Becouse I haven't idea what I can do better (database files have about 26GB size) I choose the second option.
|PBONE has indexed over 1.5 million RPM files. That's why it is so slow now. I've tried balance DB files across two raid systems. It should speed up queries processing. I work over PBONE YUM repository.
|I replace old mobo, cpu and memory modules in ftp server. It's real pleasure to see how it works now. Before ftp server had even load 100 !. Now generally it has load below 1. I also add my first 1TB hard disk to ftp server. Now I have over 4 TB disk space.
|I bougth 500GB disk becouse Mandriva mirror has bigger sizer then my biggest disk.I had problem with aircondition so ftp server was overheat and it was down for all day.
|I bought three 320GB disks to build RAID5 software array. After initial build I tried add
next disk to array by raidreconf but after 20 hours of work it failed. I lost 300GB data. Now mirror and rsync
are finishing work. I decide to run this tree disks in JBOD mode :). Now I have 3.1TB disk space.
|I had problem with ftp server. Finnaly it was problem with Antec PSU. It's third Antec PSU which failed. I'll never buy Antec PSU in the future.
|I have problem with extra 4GB RAM. Database server is reloading from time to time
|In ftp server next hdd gone wild. This time it was 120GB hdd. I've replaced it by 200GB hdd which come back from service. SCSI hdd was broken in database server too. Thanks to RAID it's not a tragedy. Database server has high load (from 5 to 20) that's why some queries are killed. I add 4GB RAM to database server. I hope it solve problem for some time.
|I buy new 300GB HDD. I spent collected money by users and I add the rest. I copy data from corrupted 200GB HDD. Now I'm able to return it and got new replacement. Mandriva mirror now is up to date. Due to high load database server I run squid as http accelerator. That's why You have trouble with language versions. I solve problem by allow only one language version. English of course. This situation will take place till I'll use squid. I bought RAID controller 3Ware to improve ftp server stability. I use JBOD mode. The controller is working very well. Now I don't have problems with DMA timeout. I've a lot of DMA timeouts when disks were connected to on board controllers.
|At the end of October I had problem with power supplier in ftp server. After 40 days of waiting I got working replacemant. It was two weeks ago. Since then I have no problems with ftp server. Except that Mandriva mirror is too big for 200GB hard disk.
|I've some problems with hard disks in ftp server. First one Seagate hard disk 80GB has gone down after 3 years and 2 months. I've repaced it by WD 320GB. Two weeks later next Seagate hard disk 120GB has gone mad after 2 years and 2 months. I didn't wait for total crash and I've replaced it by 300GB hard disk. Now I've 1.6TB disk space and I haven't any Seagate disk in pbone. I hope that is enough for next couple of months.
|Today I've brought computer which provides FTP service back to life. Kernel did oops. I've added next hard disk (320GB). Now there is about 1.1TB useful space for ftp. How nice looks 'TB'. Unfortunately I'm not able to close computer and disk is working outside chassis. Now I've 8 hard disks. Some time ago I read that kernel.org archive has 24TB. When I read something like that I'm jealous. But they have big sponsor like HP. PBone is self-sponsor. It can't be compared. It's good that ATM and 3S.PL are providing bandwidth and coolocation. Without such help even if I'll have 24TB disk space it won't be useful.
|Ftp server is down. I'll bring it back on Wensdey
|For couple of days database is operating on new server. Search time should be shorter and using website should be more comfortable. I've added next disk to ftp server. Now I've about 900GB. 820GB is already used. When I finish cleaning ftp server I'll work on my TODO list.
|Next disk in pbone. Now I've 810GB total disk space. Thanks to operator 3S network and operator ATMAN network server has new better connection to internet. New faster server for database is comming and you will see difference soon.
|I've added support for Fedora Core 3.
|Next new disk in pbone.net . Unfortunately I had to pull out 120GB disk becouse it has some bad sectors. Now I've 720GB total. When broken disk will return from service I'll replace 80GB disk with 120GB disk and I'll have 760GB total.
|I add new disk to server but I had to get old one becouse there isn't place for new disks. Now I've 640GB. I wonder how long it will be enough
|I've added support for Fedora Core 2 finally.
|Axel Thimm helped me and now you can use german language version
|Dag Wieers sold me on adding RHEL2.1 and RHEL3 support. I've also added Mandrake 10.0 support.
|I've reorganized database indexes. It helps, now server load is 1. I approve such load. I've fixed bug on NEWS page. Thanks Michal. I wonder why nobody drop me info about bug. I suspect that nobody is using customization possibility NEWS page. Strange.
|I've optimized database and I've activated SCSI disks. Becouse of that server load is three times lower. From time to time there are slow queries but now it's better than month ago. I notice some RPMs have wrong summary and description. I should insert again information about all rpms to database to fix this problem. I'll do it in near future. By the way I'll widen website functionality. Marcin has prepared new web design. When I'll have a little time I'll use it.
|I've added support for SuSE 9.X. I've removed some old distros. Now rpm files for these old distros are assigned to Other category. I've had to fix one table in database. This table holds information about rpm content. Soon (I hope) you'll see Search speed up. Unfortunately nobody offered smallest aid for server upgrade. It's a pity. Probably I were too naive.
|I've had to fix one table in database. That's why FTP search didn't work for couple of hours. I've added support for Fedora Core and ASPLinux.
|I've started collect money for new hardware. If someone is able to be a PBone sponsor (by money or by hardware) please contact with me by this form. I need motherboard which supports at least 4GB RAM, 4GB RAM, processor at least 2.8GHz (Pentium) or AMD Opteron, |SCSI U320 controler (I've bought it) two 36GB SCSI U320 hard disks at least 10kRPM (I've bought they) two 120GB IDE disks (I've already three in my server) to create RAID for RPM archive You could be RAM sponsor or one hard disk sponsor. Every help will be appreciated.
|As everyone see server is working at full capacity. I add some new sites to RPM mirror.
|I've still problems with hard disks. I had to stop adding new rpms to database. I'm waiting for new disks. I hope I'll get one tomorrow and I'll start adding new rpms.
|I've added possibility change sort search results. I've changed way of splitting search results. I hope it will help navigate. One week ago I've discovered that my problems source was broken hard disk. For that reason I had a lot of troubles. Effect of domino. Now it's all right.
|Probably I've problems with hardware. I got many different errors so it may be also kernel problem. I've checked a few kernels but I've still problems. For now server looks stable but one disk has been disconneted becouse kernel aren't able to find filesystem on it.
|Long time without diary. Couple important things take place. I upgrade operating system on server. I've still some trouble with it. Some software doesn't want to run on new redhat. Already some problems are solved but I have one very important with adding new RPMs to database. I replace 40GB hard disks by 120GB ones. My problems with free space have ended. We'll see for how long :)
|I've fixed little bug in search results. Thanks francis jones. Additional small changes.
|Next Advanced Search extension. Now you are able to choose how many results are displayed per page. I've fixed few HTML code. It is almost fully compatible with W3C standard. Busy day.
|I've fixed little bug in search results. Thanks Hellmasker.
|I've remark for these people which send me information that someting is not working. Please give your email. I'll be able to explain ambiguity or ask for more informations.
|Now you can use FTP Search with new look. I encourage you to vote which one looks better.
|I decided to remove Sourceforge, Freshmeat and Linuxberg category from Advanced Search. These categories aren't connected to any distribution.
|Litte changes on some pages. I hope that will improve functionality.
|From now pbone is ARKLinux mirror.
|From now pbone is Penguin Liberation Front full mirror.
|I started writing this diary. Here you will find what I did, do and will do.|
I merged two tables in pbone database. I hope pbone will be a little faster. I've changed way of display vendor and packager email due to potential use it by spamers. Some time ago I've fixed pbone scripts to not delete information about rpms which were deleted from my ftp server.
I always look for new rpms which I haven't in my database yet.