Creation of osd with journaling fails when no space is left on journaling device with no error message displayed
alienn
37 Posts
December 18, 2018, 3:29 pmQuote from alienn on December 18, 2018, 3:29 pmHi,
today I built a new petasan cluster and wasn't aware of the (new) 60gb partition size per osd on the journaling device.
After the 4th osd created I tried to create a fifth osd. But everytime I came back to the overview showing four osds and no error message is displayed.
After tinkering around I figured out what the problem was and was able to correct the value of bluestore_block_db_size.
I would have been nice if there had been an error message pointing me in the right direction. 🙂
Cheers,
Hi,
today I built a new petasan cluster and wasn't aware of the (new) 60gb partition size per osd on the journaling device.
After the 4th osd created I tried to create a fifth osd. But everytime I came back to the overview showing four osds and no error message is displayed.
After tinkering around I figured out what the problem was and was able to correct the value of bluestore_block_db_size.
I would have been nice if there had been an error message pointing me in the right direction. 🙂
Cheers,
yangsm
31 Posts
December 19, 2018, 12:46 pmQuote from yangsm on December 19, 2018, 12:46 pmIn bluestore_block_db_size file is the size of the log file changed?
In bluestore_block_db_size file is the size of the log file changed?
admin
2,930 Posts
December 19, 2018, 3:38 pmQuote from admin on December 19, 2018, 3:38 pmGood point, we will try to fix this in next version. fyi in these cases we do not get a failure return from ceph this is why we do not display an error, but we can check the remaining size ourselves, which we will do.
Thanks.
re bluestore_block_db_size, yes we increased this based on performance recommendations.
Good point, we will try to fix this in next version. fyi in these cases we do not get a failure return from ceph this is why we do not display an error, but we can check the remaining size ourselves, which we will do.
Thanks.
re bluestore_block_db_size, yes we increased this based on performance recommendations.
Last edited on December 19, 2018, 3:42 pm by admin · #3
Ammar
1 Post
January 3, 2019, 2:44 pmQuote from Ammar on January 3, 2019, 2:44 pmI can confirm the issue @alienn has reported. A hint to the users to specify their desired journal's slice size would be great.
Meanwhile, there are two ways to get around this...
1- During the initial installation process, when you get to Step 5;
- Choose your desired Cluster Size, Number of Replicas, and Tuning Template then click the [+ Show Details]
- Scroll to the [bluestore_block_db_size = 64424509440] , edit the value to reflect your desired slice size and continue with the installation. The value you set here will be carried over to the installed system. Tested. Works fine. Please check attached screenshot.
2- Skip adding OSDs during the initial installation process then;
- SSH into your nodes and;
-
nano +27 /opt/petasan/config/etc/ceph/PetaSAN.conf (replace PetaSAN.conf with your adequate config file name)
- change bluestore_block_db_size = 64424509440 to bluestore_block_db_size = your.preferred.value
For a 20GB slice you may use 21474836480 and so on.
- Save, reload, log-in and start adding your OSDs. Also tried and tested.
@PetaSAN team: Great work on your behalf. Keep it going.
Cheers,
Ammar
I can confirm the issue @alienn has reported. A hint to the users to specify their desired journal's slice size would be great.
Meanwhile, there are two ways to get around this...
1- During the initial installation process, when you get to Step 5;
- Choose your desired Cluster Size, Number of Replicas, and Tuning Template then click the [+ Show Details]
- Scroll to the [bluestore_block_db_size = 64424509440] , edit the value to reflect your desired slice size and continue with the installation. The value you set here will be carried over to the installed system. Tested. Works fine. Please check attached screenshot.
2- Skip adding OSDs during the initial installation process then;
- SSH into your nodes and;
-
nano +27 /opt/petasan/config/etc/ceph/PetaSAN.conf (replace PetaSAN.conf with your adequate config file name)
- change bluestore_block_db_size = 64424509440 to bluestore_block_db_size = your.preferred.value
For a 20GB slice you may use 21474836480 and so on.
- Save, reload, log-in and start adding your OSDs. Also tried and tested.
@PetaSAN team: Great work on your behalf. Keep it going.
Cheers,
Ammar
Creation of osd with journaling fails when no space is left on journaling device with no error message displayed
alienn
37 Posts
Quote from alienn on December 18, 2018, 3:29 pmHi,
today I built a new petasan cluster and wasn't aware of the (new) 60gb partition size per osd on the journaling device.
After the 4th osd created I tried to create a fifth osd. But everytime I came back to the overview showing four osds and no error message is displayed.
After tinkering around I figured out what the problem was and was able to correct the value of bluestore_block_db_size.
I would have been nice if there had been an error message pointing me in the right direction. 🙂
Cheers,
Hi,
today I built a new petasan cluster and wasn't aware of the (new) 60gb partition size per osd on the journaling device.
After the 4th osd created I tried to create a fifth osd. But everytime I came back to the overview showing four osds and no error message is displayed.
After tinkering around I figured out what the problem was and was able to correct the value of bluestore_block_db_size.
I would have been nice if there had been an error message pointing me in the right direction. 🙂
Cheers,
yangsm
31 Posts
Quote from yangsm on December 19, 2018, 12:46 pmIn bluestore_block_db_size file is the size of the log file changed?
In bluestore_block_db_size file is the size of the log file changed?
admin
2,930 Posts
Quote from admin on December 19, 2018, 3:38 pmGood point, we will try to fix this in next version. fyi in these cases we do not get a failure return from ceph this is why we do not display an error, but we can check the remaining size ourselves, which we will do.
Thanks.
re bluestore_block_db_size, yes we increased this based on performance recommendations.
Good point, we will try to fix this in next version. fyi in these cases we do not get a failure return from ceph this is why we do not display an error, but we can check the remaining size ourselves, which we will do.
Thanks.
re bluestore_block_db_size, yes we increased this based on performance recommendations.
Ammar
1 Post
Quote from Ammar on January 3, 2019, 2:44 pmI can confirm the issue @alienn has reported. A hint to the users to specify their desired journal's slice size would be great.
Meanwhile, there are two ways to get around this...
1- During the initial installation process, when you get to Step 5;
- Choose your desired Cluster Size, Number of Replicas, and Tuning Template then click the [+ Show Details]
- Scroll to the [bluestore_block_db_size = 64424509440] , edit the value to reflect your desired slice size and continue with the installation. The value you set here will be carried over to the installed system. Tested. Works fine. Please check attached screenshot.
2- Skip adding OSDs during the initial installation process then;
- SSH into your nodes and;
nano +27 /opt/petasan/config/etc/ceph/PetaSAN.conf (replace PetaSAN.conf with your adequate config file name)
- change bluestore_block_db_size = 64424509440 to bluestore_block_db_size = your.preferred.value
For a 20GB slice you may use 21474836480 and so on.
- Save, reload, log-in and start adding your OSDs. Also tried and tested.
@PetaSAN team: Great work on your behalf. Keep it going.
Cheers,
Ammar
I can confirm the issue @alienn has reported. A hint to the users to specify their desired journal's slice size would be great.
Meanwhile, there are two ways to get around this...
1- During the initial installation process, when you get to Step 5;
- Choose your desired Cluster Size, Number of Replicas, and Tuning Template then click the [+ Show Details]
- Scroll to the [bluestore_block_db_size = 64424509440] , edit the value to reflect your desired slice size and continue with the installation. The value you set here will be carried over to the installed system. Tested. Works fine. Please check attached screenshot.
2- Skip adding OSDs during the initial installation process then;
- SSH into your nodes and;
-
nano +27 /opt/petasan/config/etc/ceph/PetaSAN.conf (replace PetaSAN.conf with your adequate config file name)
- change bluestore_block_db_size = 64424509440 to bluestore_block_db_size = your.preferred.value
-
For a 20GB slice you may use 21474836480 and so on.
- Save, reload, log-in and start adding your OSDs. Also tried and tested.
@PetaSAN team: Great work on your behalf. Keep it going.
Cheers,
Ammar