Root disk size reverts to its template size after reinstall

This issue is almost the same as the #3957

ISSUE TYPE

Bug Report

COMPONENT NAME

Instance reinstallation

CLOUDSTACK VERSION

main,
4.17, 
but not in 4.14

SUMMARY

Reinstall action acts surprisingly. When users reinstall a vm, they expect the VM configurations stay the same (like: service offering, volume sizes, IP) and a brand new operating system installs on the VM. But it is not always true. For example:

First Scenario:

Deploy an instance. Don't override the root disk
Resize the root volume
Reinstall the VM

Expected Result	Actual Result
Root volume size stays the same	Root volume size resets to original value

First Scenario:

Deploy an instance. Override the root disk
Resize the root volume
Reinstall the VM

Expected Result	Actual Result
Root volume size stays the same	Root volume size stays the same

Findings

In our debugging we noticed that when user overrides the root disk size an entry in user_vm_details is created. When user don't override root disk size, the entry is not created in database. In that case, CloudStack uses the template's disk size rather than the old root volume size in reinstallation process.

mysql> select * from user_vm_details where name = 'rootdisksize' ;
+----+-------+--------------+-------+---------+
| id | vm_id | name         | value | display |
+----+-------+--------------+-------+---------+
| 21 |     5 | rootdisksize | 15    |       1 |
+----+-------+--------------+-------+---------+
1 row in set (0.00 sec)

cloudstack/server/src/main/java/com/cloud/vm/UserVmManagerImpl.java

Lines 7746 to 7752 in eac357c

    
           if (userVmDetailsDao.findDetail(vm.getId(), VmDetailConstants.ROOT_DISK_SIZE) == null && !newVol.getSize().equals(template.getSize())) { 
        
               VolumeVO resizedVolume = (VolumeVO) newVol; 
        
               if (template.getSize() != null) { 
        
                   resizedVolume.setSize(template.getSize()); 
        
                   _volsDao.update(resizedVolume.getId(), resizedVolume); 
        
               } 
        
           }

fixed by #7145

@weizhouapache

@weizhouapache

Although the #7145 PR fixed the issue for new VMs, we have to update CloudStack's database manually and add an entry in user_vm_details for old VMs. I thought it would be nice to share the scrips here just in case someone has the same issue. :-)

Get the affected VMs with the following sql query. Save it to a file called missing.txt
Then use the python code to generate appropriate insert SQLs. python3 parser.py missing
The python programm will generate the missing.sql file.
Update the database using the missing.sql

MariaDB [cloud]> SELECT distinct(view.id) as vm_id, view.uuid as vm_uuid, view.volume_id, vol.uuid as volume_uuid, \
    vol.volume_type , details.value as rootdisksize, CAST(vol.size/1024/1024/1024 AS SIGNED INTEGER) as expected \ 
    FROM cloud.user_vm_view AS view LEFT JOIN cloud.user_vm_details AS details ON view.id=details.vm_id AND \
    details.name='rootdisksize' LEFT JOIN cloud.volumes AS vol ON vol.id=view.volume_id WHERE view.state != 'Destroyed' \
    AND (CAST(vol.size/1024/1024/1024 AS SIGNED INTEGER) != details.value OR details.value IS NULL) \
    and vol.volume_type = 'ROOT' and details.value is NULL ORDER BY vm_id ;
+-------+--------------------------------------+-----------+--------------------------------------+-------------+--------------+----------+
| vm_id | vm_uuid                              | volume_id | volume_uuid                          | volume_type | rootdisksize | expected |
+-------+--------------------------------------+-----------+--------------------------------------+-------------+--------------+----------+
|   1   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff0 |       1   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff0 | ROOT        | NULL         |       12 |
|   2   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff1 |       2   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff1 | ROOT        | NULL         |       15 |
|   3   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff2 |       3   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff2 | ROOT        | NULL         |       25 |
+-------+--------------------------------------+-----------+--------------------------------------+-------------+--------------+----------+
3 rows in set (0.005 sec)

import sys

input_file_name = '{}.txt'.format( sys.argv[1] )
output_file_name = '{}.sql'.format( sys.argv[1] )

with open( input_file_name , "r") as f_in:
    with open( output_file_name , "w") as f_out:
        for li in f_in.readlines():
            if '-+-' not in li and "vm_id" not in li and "SELECT" not in li and 'rows' not in li:
                print(li, end = '')
                tmp = li.split('|')
                print(tmp)
                out_line = 'insert into user_vm_details (vm_id, name, value, display) values ( {}, \'rootdisksize\', {}, 1);\n'.format( tmp[1], tmp[-2] )
                print(out_line, end = '')

                if 'NULL' in tmp[-3]:
                    f_out.write(out_line)

@weizhouapache

@weizhouapache

Although the #7145 PR fixed the issue for new VMs, we have to update CloudStack's database manually and add an entry in user_vm_details for old VMs. I thought it would be nice to share the scrips here just in case someone has the same issue. :-)

Get the affected VMs with the following sql query. Save it to a file called missing.txt
Then use the python code to generate appropriate insert SQLs. python3 parser.py missing
The python programm will generate the missing.sql file.
Update the database using the missing.sql

MariaDB [cloud]> SELECT distinct(view.id) as vm_id, view.uuid as vm_uuid, view.volume_id, vol.uuid as volume_uuid, \
    vol.volume_type , details.value as rootdisksize, CAST(vol.size/1024/1024/1024 AS SIGNED INTEGER) as expected \ 
    FROM cloud.user_vm_view AS view LEFT JOIN cloud.user_vm_details AS details ON view.id=details.vm_id AND \
    details.name='rootdisksize' LEFT JOIN cloud.volumes AS vol ON vol.id=view.volume_id WHERE view.state != 'Destroyed' \
    AND (CAST(vol.size/1024/1024/1024 AS SIGNED INTEGER) != details.value OR details.value IS NULL) \
    and vol.volume_type = 'ROOT' and details.value is NULL ORDER BY vm_id ;
+-------+--------------------------------------+-----------+--------------------------------------+-------------+--------------+----------+
| vm_id | vm_uuid                              | volume_id | volume_uuid                          | volume_type | rootdisksize | expected |
+-------+--------------------------------------+-----------+--------------------------------------+-------------+--------------+----------+
|   1   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff0 |       1   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff0 | ROOT        | NULL         |       12 |
|   2   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff1 |       2   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff1 | ROOT        | NULL         |       15 |
|   3   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff2 |       3   | aaaaaaaa-bbbb-cccc-dddd-fffffffffff2 | ROOT        | NULL         |       25 |
+-------+--------------------------------------+-----------+--------------------------------------+-------------+--------------+----------+
3 rows in set (0.005 sec)

import sys

input_file_name = '{}.txt'.format( sys.argv[1] )
output_file_name = '{}.sql'.format( sys.argv[1] )

with open( input_file_name , "r") as f_in:
    with open( output_file_name , "w") as f_out:
        for li in f_in.readlines():
            if '-+-' not in li and "vm_id" not in li and "SELECT" not in li and 'rows' not in li:
                print(li, end = '')
                tmp = li.split('|')
                print(tmp)
                out_line = 'insert into user_vm_details (vm_id, name, value, display) values ( {}, \'rootdisksize\', {}, 1);\n'.format( tmp[1], tmp[-2] )
                print(out_line, end = '')

                if 'NULL' in tmp[-3]:
                    f_out.write(out_line)

thanks @soreana for sharing
would it be good do it via mysql procedure?

@weizhouapache

@weizhouapache We had a lot of edge cases, we prefer to get the list first and manually update those. :-)

Navigation Menu

Repository navigation

Root disk size reverts to its template size after reinstall #7139

ISSUE TYPE

COMPONENT NAME

CLOUDSTACK VERSION

SUMMARY

First Scenario:

First Scenario:

Findings

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	if (userVmDetailsDao.findDetail(vm.getId(), VmDetailConstants.ROOT_DISK_SIZE) == null && !newVol.getSize().equals(template.getSize())) {
	VolumeVO resizedVolume = (VolumeVO) newVol;
	if (template.getSize() != null) {
	resizedVolume.setSize(template.getSize());
	_volsDao.update(resizedVolume.getId(), resizedVolume);
	}
	}

Root disk size reverts to its template size after reinstall #7139

Description

ISSUE TYPE

COMPONENT NAME

CLOUDSTACK VERSION

SUMMARY

First Scenario:

First Scenario:

Findings

Activity

DaanHoogland commented on Feb 1, 2023

soreana commented on Feb 10, 2023

weizhouapache commented on Feb 11, 2023

soreana commented on Feb 16, 2023

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions